Reprinted from The Official Journal of ISPE PHARMACEUTICAL ... · mization can be executed not only in a scale-up, but also in a scale-down experiment for an existing fermenter. Downstream

Process Modeling

NOVEMBER/DECEMBER 2005 PHARMACEUTICAL ENGINEERING 1©Copyright ISPE 2005

Process Modeling Proposition inBiopharmaceutical Manufacturingby Sei Murakami, PhD, Peter Watler, PhD,Takashi Ishihara, PhD, and Shuichi Yamamoto, PhD

This articledescribesadvancementsin processmodeling inbiopharmaceuticalmanufacturingfocusing onfermentationandchromatographicseparation.

Table A. Empiricalequations for similarfigures scale-up.

Characteristics Empirical Equations 200 L 10,000 L

Power Input Volumetric Agitation Pg/V n3 Di5 1.00 1.00 0.20 0.27 13.6Power ∝ ______

V

Mass Transfer Volumetric Oxygen kLa Pg 1.00 1.92 1.00 1.14 5.45Transfer Coefficient ∝ ___ 0.4

Us0.5

V

Hydrodynamic Shearing Rate by (dU/dz)max ∝ nDi 1.00 1.54 0.90 1.00 3.68Intensity Impeller Tip Speed

Homogeneity Circulation Rate Qi/V n Di3 1.00 0.42 0.24 0.27 1.00∝ ______

V

Introduction

Process modeling and simulation areexpected to enable the identificationand evaluation of product and processvariables that may be critical to prod-

uct quality and performance. They also mayidentify potential failure modes and mecha-nisms and quantify their effects on productquality before and during the actual process-ing.

However, in spite of those benefits, processmodeling for biopharmaceutical manufactur-ing has not been developed extensively due toits reaction and molecular structure complex-ity. In order to contribute to the process under-standing in the biopharmaceutical industry,advancements in process modeling inbiopharmaceutical manufacturing focusing onfermentation as an upstream and chromato-graphic separation as a downstream represen-tative unit operation will be described.

Upstream - FermentationLimitations in Similar Figures Scale-UpLarge-scale mammalian cell culture forbiopharmaceutical production is always intri-cate due to mammalian cell’s extreme fragilityand complex nutrient requirement. Avoidingcell damage, while optimizing oxygen supply

and carbon dioxide extraction, is the key to themammalian cell culture scale-up. In order tocharacterize such fermentation conditions forpredicting scale-up results, various empiricalequations have been introduced. Since most ofsuch equations are dependent on the figures,there needs to be similarity between experi-mented equipment and scaled-up fermenter.Accordingly, fermentation scale-up has beenaccomplished with similar figures in order tokeep reliability of such empirical equationsusage. However, all fermentation characteris-tics cannot be kept constant simultaneouslyduring the similar figures scale-up.1 For ex-ample, when we scale-up a fermenter from 200to 10,000 liters with constant impeller tip speed,volumetric power input decreases to almost onequarter - Table A.

Accordingly, the actual manufacturing recordfor the cell culture fermenter shows a rapiddecline in volumetric power input during scale-up, resulting from a restriction of constantimpeller tip speed - Figure 1.

Model DevelopmentTo facilitate a more flexible scale-up with quan-titative culture environment predictions, a fer-menter model independent of figures similar-ity needs to be provided. Modeling of a fer-

( )

Reprinted from The Official Journal of ISPE

PHARMACEUTICAL ENGINEERING® November/December 2005, Vol. 25 No. 6

Process Modeling

2 PHARMACEUTICAL ENGINEERING NOVEMBER/DECEMBER 2005 ©Copyright ISPE 2005

menter requires addressing the following problems: turbu-lent flow, gas-liquid multi-phase flow, and moving boundaryconditions of impellers and baffles.

In addition to using a direct solving method of the three-dimensional Navier-Stokes equation, turbulent flow has beenaddressed utilizing the eddy-viscosity model, the Reynoldsstress equation model, the large eddy simulation, and thevortex method. Among them, a subset of the eddy-viscositymodel, k-ε model,3 has been widely used due to its wideapplicability and a moderate computational resource re-quirement. In spite of limitations of the k-ε model, such asdependence of six empirical constants and an isotropic turbu-lence assumption, its applicability to mixing vessel simula-tion has been confirmed and excellent representation ofexperimentally measured values, especially flow velocity,have been reported.4

Modeling of multi-phase flow includes the Eulerian andLagrangian models. The Eulerian model assumes each phaseto have a separate velocity field and a common pressure field,whereas the Lagrangian model tracks representative bubblesthrough the domain. A large number of bubbles often ob-served in fermenter operation renders it unrealistic to applythe Lagrangian model’s bubble tracking. A simpler subset ofthe Eulerian multi-fluid model is drift-flux model,5 whichassumes the dispersed phase, bubbles, move relative to the

continuous phase, liquid, at their terminal velocity. In thedrift-flux model, bubble acceleration and physical propertychange, usually not significant in fermentation, are ne-glected. With gas hold up, bubble diameter, and hydrody-namic parameters, a volumetric oxygen transfer coefficientkLa can be described. Consequently, dissolved oxygen andcarbon dioxide distribution can be calculated by solvingoxygen and carbon dioxide transport equations.

If a fermenter vessel does not have baffles, boundaryconditions can be simpler and do not need to be changed withtime by defining it rotating with impellers. However, mostfermenters are furnished with baffles or other internal fixedstructures making the simulation model’s boundary condi-tion multifarious. In order to avoid the complexity of timedependent boundary conditions, a method with an equivalentcylindrical block having experimentally determined bound-ary conditions of flow velocity and turbulence representingthe rotating impellers has been introduced. Although it issimple to calculate, an impeller representative cylinder al-ways requires actual scale experimental measurement andcannot be used for unrealized scale-up study. To avoid theempirical dependence, the dynamical multi-block method6

and the sliding mesh method7 have been introduced whereimpellers and baffles are modeled in a separate block rotatingrelatively to each other.

One of the descriptions of hydrodynamic damage to thecultured cells is Kolmogoroff Eddy Length Scale, which isassumed to be the minimum eddy size before dissipated byviscous force. If the Kolmogoroff Eddy Length Scale is smallerthan that of solid particle such as suspended cells, theparticle may not move to release the hydrodynamic force andmay receive some damage on its surface. On the other hand,if the Eddy is larger, the particle can float on the eddy, andthus, hydrodynamic damage can be avoided.8

Based on the above modeling, empirical equations forfermentation characteristics shown in Table A can be substi-tuted by the following elemental and overall parametersindependent of figures similarity among different scales.

Because these parameters are independent of reactorfigures, they can be used for various fermenter shapes.

SimulationUsing models described above, one now can perform fermen-

Figure 1. Volumetric power input of various scale fermenters.2

( )

Table B. Elemental parameters and overall evaluation for fermentation model.

Characteristics Elemental Parameters Overall Evaluation

Power Input Turbulent Energy ρε Total Turbulent Energy 1Dissipation Rate Dissipation Rate ____ ∫∫∫ (ρε) dxdydz V

Mass Transfer Local kLa (α/Db) f (Sc, ν, k, ε) Total kLa 1____ ∫∫∫ (kLa) dxdydz V

Hydrodynamic Intensity Kolmogoroff Eddy (ν3/ε)1/4 Minimum Eddy Length Scale ν3 1/4

Length Scale ____ε min

Homogeneity Local Concentration C Standard Deviation of 1 —Concentration ____ ∫∫∫ (C–C)2 dxdydz√ V

Process Modeling


tation environments simulations, which provide liquid flowvector, gas hold up, hydrodynamic damage, volumetric oxy-gen transfer coefficient, kLa, dissolved oxygen, dissolvedcarbon dioxide, etc. Due to culture liquid physical character-istics that cannot be predicted precisely prior to the simula-tion, primary simulation results might differ from experi-mental results. Consequently, appropriate adjustment for aparticular fermentation is essential. This adjustment can bedone by performing a couple of simulations for establishedbench or small-scale fermentations, then comparing thesimulation and experimental results.

Process OptimizationHydrodynamic damage, dissolved oxygen, and dissolved car-bon dioxide are the key critical fermentation conditions formammalian cell culture. The following discussion describessome of the applications of the proposed model and simulationfor evaluating and optimizing a fermentation environment.

As a hydrodynamic intensity evaluation example,Kolmogoroff Eddy Length Scale distributions are shown inFigure 2. Depending on the size of the particle, such as asuspended cell and a micro carrier, appropriate impeller typeand agitation speed, which has larger Eddy Length Scale, canbe found by changing the impeller profile in the simulation.

The dissolved carbon dioxide that is often increased byscale-up will likely reduce cell growth and productivity.Because liquid surface mass transfer has a significant effecton carbon dioxide extraction, a fermenter aspect ratio needsto be optimized. Figure 3 shows an effect of bioreactor aspectratio on dissolved carbon dioxide reduction.

As described above, flexible fermenter configuration andoperating conditions evaluations provide minimum hydrody-namic damage and dissolved carbon dioxide which lead toproductivity maximization. This simulation for process opti-mization can be executed not only in a scale-up, but also in ascale-down experiment for an existing fermenter.

Downstream - Chromatographic SeparationChromatography Processes forBiopharmaceutical ProductsSeparation and purification processes of recombinant pro-tein-based pharmaceuticals and other biological productsinvolve a series of chromatography steps. Highly advancedsimulation software is available for separation unit opera-tions in chemical and petrochemical processes such as distil-lation. This type of software is now routinely used for opera-tion and optimization of processes as well as process design.However, chromatography processes are still designed andoperated on the basis of the knowledge empirically obtainedor on the trial-error approaches. Therefore modeling andsimulation of chromatography are important for rationaldesign, optimization, and stable operation of processes.

Several modes of chromatography are available to exploitthe charge, hydrophobicity, and size differences of the con-taminants and the product. In addition, chromatographycolumns can be operated in various modes such as flow-through or adsorption mode with isocratic or gradient elu-

tion. Current chromatography column technology has en-abled the size of preparative columns from a few centimetersto up to two meters capable of holding hundreds of liters ofpacking materials.

To take advantage of these many advancements, it iscritical to have a well-planned development strategy andsolid understanding of the separation mechanism to success-fully design, scale-up, and implement a cost effective chro-matographic process. Indeed, this ability is a key factor in thesuccess of a company’s process development effort and subse-quent product commercialization.

When a certain chromatography process can be predictedby a model simulation, it is easy to optimize the process bytuning the operating variable and designing a better perfor-mance process. Plant constraints such as buffer and tankvolumes and process time also can be incorporated into themodel. The model allows rapid assessment of buffer usage,retention times, peak volumes, and estimates purity, recov-ery, and productivity. Also it is easy to change the conditionsto observe how the separation changes. The model provides asolid understanding of the process, which parameters areimportant, and how and why they affect the separation.9-12

The following discussion will describe how to apply modelsto chromatography processes and what can be done withmodels.9

Figure 3. Dissolved carbon dioxide distribution.

Figure 2. Kolmogoroff Eddy Length Scale distribution.

Process Modeling


Use of Model forScale-Up to Industrial Size ColumnChromatography peaks are characterized with the retentionvolume and the peak width. The retention volume is relatedto the distribution coefficient K, and the peak width isexpressed by HETP = Z/N (Z: column bed height, N: platenumber).9-12 The model parameters K as a function of saltconcentration I, and HETP as a function of velocity aredetermined from a series of small-scale Linear GradientElution (LGE) experiments.9,13 These data also can be appliedto isocratic elution. As shown in Figure 4, the model can beused to predict isocratic elution behavior of a large-scaleindustrial column. The predicted chromatogram closely ap-proximated the actual chromatogram from a 40 L columnwith predicted retention volumes within ~15% of the actualretention volume.

Use of Model for Linear GradientElution OptimizationOne of the most useful applications of a chromatographymodel is for screening operating conditions to characterizeand optimize the separation. To assess the separation andchallenge the model, extremes of column length, gradientlength, and initial ionic concentration were evaluated.

A tall bed height with a shallow gradient slope typicallymaximizes peak separation, but also increases the separa-tion time and the elution volume. Such conditions werescreened with the aid of the model by adjusting the operatingconditions to a 20 cm bed height and 30 (Column Volume) CVgradient. Under these conditions, the model predicts rela-tively broad and late eluting peaks with baseline. To verifythe model, the actual chromatogram obtained under thesame operating conditions is shown in Figure 5. The pre-dicted and actual chromatograms show similar peak shapesand peak widths, and baseline separation with the main peakeluting at ~70% of the gradient. The model provided anaccurate prediction of the separation under conditions forhigh resolution.

After identifying and verifying a high resolution separa-tion, the operating conditions were modified to search forconditions that gave good resolution, but were more scaleable

and offered higher productivity. A shorter bed height willhave lower pressure drop upon scale-up, and a more moder-ate gradient length will lower buffer consumptions. To com-pensate for the shorter gradient length, gradient slope wasreduced by increasing the initial ionic concentration, I0. Thepredicted chromatogram from the more moderate operatingconditions of 12 cm bed height, 20 CV gradient, and I0 = 28mM showed near baseline separation of the peaks with theproduct peak eluting earlier at ~55% of the gradient and witha narrower peak width9 - separation was verified experimen-tally, showing that the LGE model provided an accurateprediction under moderate operating conditions.9

Finally, the separation behavior under extremely lowresolution conditions was studied. A very low bed height (5cm) and a very short gradient (7 CV) required minimalpacking materials, buffer and tanks. However, the modelpredicted that the pre-peaks merged with the product peak,appearing as a small shoulder on the product peak resultingin very poor resolution.9 In addition, the model predicted thatthe product peak eluted toward the end of the gradient(~90%). The actual chromatogram obtained at these operat-ing conditions was very similar to the predicted chromato-gram, showing a very sharp, late eluting peak with littleresolution.

As discussed above, chromatography models can be usedto screen a wide range of conditions in order to optimize andcharacterize the separation. It is important to empiricallyverify the proposed conditions prior to specifying them forscale-up. Table C shows that the model predicted how theelution volume and peak width changed at various columnand gradient conditions. Predicted retention volume waswithin 3% of the actual volume for the product peak. Themodel showed that peak width decreased with sharper gradi-ents and lower bed heights, but under-predicted values by 21-41%.

Use of Model to Improve ProductivityProductivity is defined as the amount of protein of a givenpurity, produced per unit time per liter of chromatographypacked bed volume. Models can be employed to rapidly scoutconditions which give the highest productivity for a givenpurity requirement. For the separation shown in Figure 4,the optimized isocratic operating condition gave a produc-tivity of 0.16 g/L.h - Table D. This is largely because of thelong cycle time due to the high distribution coefficient of theproduct peak under isocratic elution at 40 mM NaCl. Onemethod of addressing a high K value is to use gradientelution. The model showed that a very large gradient vol-ume (60 CV) gave comparable resolution with a somewhathigher productivity of 0.21 g/L.h. From this starting point,operating conditions were varied to search for increasedproductivity. A moderate bed height of 10 cm and a moder-ate gradient of 10 CV gave sufficient resolution to achieve99% purity and very high (99%) recovery. One of the featuresof the model is that the starting ionic strength can be variedto adjust the gradient slope. Increasing the starting saltconcentration to 40 mM, resulted in almost immediate

Figure 4. Comparison of isocratic elution curve predicted from themodel using a small (9 mL) column linear gradient elution data andactual elution curve of a 40 L column: CM Sepharose FF.9

Process Modeling


Figure 5. Comparison of predicted chromatogram (a) and actualchromatogram (b) for CM Sepharose FF, Gradient = 30 CV,diameter = 1.6 cm, height = 20 cm, Io = 16mM NaCl,superficial velocity = 120 cm/h.9

desorption of the pre-peaks as was indicated by the K-Icurve. At this gradient slope, the product peak elutes earlierresulting in a short cycle time and higher productivity.Another feature of this separation is that the steeper gradi-ent results in sharper peaks and a lower product poolvolume. Productivity under these conditions was 0.61 g/L.h,nearly 400% higher than the initial optimized isocraticconditions.

Use of Model for Characterization andTroubleshooting of SeparationTo maintain process consistency, validated large-scaleprocesses typically have a constraint to maintain elutionvolume within ±5%. This is often needed as tank volumesare fixed and elution must be completed within plannedproduction times. While column size, flow-rate, and pro-tein loading are fixed by the manufacturing procedure,resin ionic capacity, buffer pH, and ionic strength will varydue to lot-to-lot variations. To assess the consistency of theseparation, the model can be used to characterize andtroubleshoot the effect of these fluctuations. In this appli-cation, the model can be used to elucidate the bindingcharacteristics and to quantify changes in elution volumewith variations of media ion-exchange capacity and bufferionic strength and the pH.14 For illustration, a modelseparation system of β-lactoglobulin near its isoelectricpoint on a weak cation exchanger, CM Sepharose at pH 5.2,was selected.14 Although the isoelectric point of this pro-tein is 5.1-5.2, it is retained on both anion and cationexchange chromatography columns at pH ~5.2.15 Accordingto the manufacturer, the ion-exchange capacity Λ of CMSepharose FF ranges from 90 to 130 mmol/mL-gel. Suchvariations can greatly affect the retention time since thedistribution coefficient is related to Λ. For the recombinantprotein separation shown in Figure 4, a 22% increase in thetotal ionic capacity from 90 to 110 mM mmol/mL resultedin a 47% increase in the distribution coefficient and therelative retention volume.9

In a production environment, it is impractical to consis-tently obtain resin of a specific ionic capacity. In order tocontrol the retention volume, it is necessary to adjust the saltconcentration IE, of the elution buffer. In the above-men-tioned model separation system with β-lactoglobulin, therelationship between the relative elution volume and IE wasexamined. It was found that the NaCl concentration must beadjusted by ±0.015 mol/L to elute the protein within ±5% of

the reference elution volume.This is typically done by trial and error. However, as

discussed above, the distribution coefficient as a function ofionic strength can be obtained from gradient elution experi-mental data. Once this K-I information is obtained for a givenΛ, the elution volume can be predicted and the elution bufferionic strength can be adjusted. Hence, the chromatographymodel can serve as a convenient tool for tuning and trouble-shooting very sensitive isocratic chromatography processes.In addition, there is usually a variation in the salt concentra-tion of elution buffers prepared at production scale. In thisexample, the salt concentration of the buffer must be within±0.002 mol/L in order to meet the ±5% elution volumecriteria.

Due to inherent variability during preparation, buffer pHalso will vary at production scale. Buffer pH affects thecharged state of the of the ion exchanger which affects elutionvolume. The ion exchange capacity of CM-Sepharose de-creases with pH below pH 6.10 The relative change in elutionvolume, resulting from changes in the ion-exchange capacity,Λ as a function of operating pH was investigated for the modelseparation system. Although variations in the relative elu-tion volume are smaller compared with the salt concentration

Operating Conditions Retention Volume (mL) Product Peak Width (mL)

Bed Height (cm) Gradient (CV) Initial Salt Predicted Actual Predicted Actual Concentration (mM)

20 30 16 164 162 71 44

12 20 28 110 110 56 42

12 50 16 230 228 --- ---

5 7 16 66 68 42 33

Table C. Comparison between predicted and actual experimental retention volume and peak width for the product peak under variousgradient elution conditions and column geometries, CM-Sepharose FF superficial velocity = 120 cm/h.9

Process Modeling


Operating ConditionsRecovery Productivity

Bed Height Gradient Io (%) (g/L.h)(cm) (CV) (mM)

12 Isocratic 40 100 0.16

12 60 0 100 0.21

20 30 16 100 0.24

5 7 16 28 0.53

10 10 40 99 0.61

Table D. Optimization of operating conditions using chromatographymodel simulations. Comparison of recovery and productivity forvarious operating conditions, CM-FF, If=100 mM NaCl.9

and the ionic-capacity, it is still important that the buffer pHbe within ±0.1 pH unit. It is additionally important to controlbuffer pH since as discussed above, the interaction betweenthe protein and the ion-exchanger changes with pH especiallynear the isoelectric point. However, controlling the effect ofpH is more complex than controlling salt concentration or theion-exchange capacity.

ConclusionAdvancements in process modeling for fermentation andchromatographic separation were described.

For a fermentation process, hydrodynamic and the masstransfer model have been incorporated into CFD, whichenables us to overcome contradictions in similar figuresscale-up. While all parameters expressing fermentationconditions cannot be kept constant simultaneously duringthe scale-up, the proposed model facilitates the scale-upenvironment prediction along with flexible fermenter con-figuration and operating conditions for productivity maxi-mization. This simulation method also can be used foranalyzing and understanding an existing fermentation pro-cess for further quality and productivity enhancement.Appropriate adjustment for a particular fermentation byperforming a couple of simulations for established bench orsmall scale fermentations is necessary for more accurateperformance prediction.

For a downstream process, the model simulation is auseful tool for process design, diagnosis and operations. Italso is helpful to understand the mechanism of very difficultand unstable separations. We have applied the model analy-sis to the separation of protein variants near the isoelectricpoints,15 the separation of monoclonal antibodies,16 and theseparation with monolithic columns.17 Further study is neededto establish a fast and simple method for determining dataneeded for the model simulations, and a method for obtainingimportant information with the aid of rapidly developing“bioinformatics.”12

Both models for upstream and downstream biophar-maceutical manufacturing described here will provide in-sight and understanding of the critical process attributes,which enable superior performance prediction, proper pro-cess monitoring interpretation, in-process adjustment, andversatile troubleshooting.

FDA’s regulatory framework (Process Analytical Tech-nology or PAT) is intended to facilitate progress to thedesired state of pharmaceutical manufacturing.18 In one ofthe PAT Tools “Multivariable tools for design, data acqui-sition, and analysis,” mathematical relationships andmodels are expected to provide scientific understanding ofthe relevant multi-factorial relationships. In conjunctionwith recent biopharmaceutical manufacturing technologydevelopment concerning other PAT Tools, i.e. “ProcessAnalyzers,” “Process Control Tools,” and “Continuous Im-provement and Knowledge Management,” these modelspossibly will contribute to the progress of the PAT frame-work.

NomenclatureC concentration for fermenter homogeneity evalua-

tion [-]CV column volume [L]Db bubble diameter [m]D diffusivity [m2/s]Di impeller diameter [m]HETP height equivalent to a theoretical plate [cm]I ionic strength of buffer [M]I0 initial ionic strength of buffer [M]IE ionic strength of elution buffer [M]K distribution coefficient [-]k turbulent energy [m/s]kLa volumetric mass transfer coefficient [s-1]n rotation speed [s-1]Pg sparged mixing power [w]Qi impeller pumping flow rate [m3/s]Sc Schmidt number [-]Sct turbulent Schmidt number [-]u velocity vector [m/s]Ui impeller tip speed [m/s]Us reactor superficial gas velocity [m/s]V reactor volume [m3]Z column bed height [cm]α gas hold up [-]ε turbulent energy dissipation rate [m2/s3]η Kolmogoroff Eddy Length Scale [m]Λ ion-exchange capacity [µmol/mL-gel]ν kinetic viscosity [m2/s]νe turbulent kinetic viscosity [m2/s]ρ density [kg/m3]

References1. Oldshue, J.Y., “Fermentation Mixing Scale-Up Tech-

niques,” Biotechnol. Bioeng., Vol. 8, 1966, pp. 3-24.2. Murakami, S., R. Nakano, and T. Matsuoka, “Scale-Up of

Fermenter: Survey of Industrial Fermenter Specifica-tions,” Kagaku Kougaku Ronbunshu, Vol. 26, 2000, pp.557-562.

3. Launder, B.E. and D.B. Spalding, “The Numerical Com-putation of Turbulent Flows,” Comp. Meth. Appl. Mech.Eng., Vol. 3, 1974, pp. 269-289.

Process Modeling


4. Ranade, V.V., “Computational Fluid Dynamics for Reac-tor Engineering,” Reviews in Chemical Engineering, Vol.11, 1995, pp. 229-289.

5. Zuber, N. and A. Findley, “Average Volumetric Concen-tration in Two-Phase Flow System,” Trans. ASME, J.Heat Transfer, Vol. 87, 1965, pp. 453-468.

6. Takeda, H. and C. Z. Hsu, “A Finite-Difference Method forIncompressible Flows Using a Multi-Block Technique,”12th International Conference on Numerical Methods inFluid Dynamics, Springer-Verlag,1990, p. 545-549.

7. Perng, C. Y. and J. Y. Murthy, “A Moving-Deforming-Mesh Technique for Simulation of Flow in Mixing Tanks,”AIChE Symp. Series, Vol. 89, Meet. Miami Beach, Fl.,1992, pp. 37-41.

8. Croughan, M.S., J. Hamel and D.I.C. Wang, “Hydrody-namic Effects on Animal Cells Grown in MicrocarrierCultures,” Biotechnol. Bioeng., Vol. 29, 1987, pp. 130-141.

9. Watler, P.D., O. Kaltenbrunner, D. Feng and S. Yamamoto,“Engineering Aspects of Ion-Exchange Chromatography”in “Scale-Up and Optimization in Preparative Chroma-tography Principles and Biopharmaceutical Applications,”eds Rathore, A.S. and Velayudhan, A., Dekker, New York,2002, p.123-171.

10. Yamamoto, S., K. Nakanishi and R. Matsuno, Ion-Ex-change Chromatography of Proteins, Marcel Dekker, 1988.

11. Guiochon, G., S.G. Shirazi and A.M. Katti, Fundamentalsof preparative and nonlinear chromatography, AcademicPress, Boston, 1994.

12. Ladisch, M.R., Bioseparations Engineering, Wiley, NewYork, 2001.

13. Yamamoto, S., “Plate Height Determination for GradientElution Chromatography of Proteins,” Biotechnol. Bioeng.,48, 5, 1995, pp. 444-451.

14. Yamamoto, S., P. Watler, D. Feng, and O. Kaltenbrunner,“Characterization of Unstable Ion-Exchange Chromato-graphic Separation of Proteins,” J. Chromatogr. A, 852,1999, pp. 37-41.

15. Yamamoto, S. and T. Ishihara, “Resolution and Retentionof Proteins Near Isoelectric Points in Ion Exchange Chro-matography - Molecular Recognition in Electro StaticInteraction Chromatography,” Sep. Sci. Tech, 35, 2000,pp. 1707-1717.

16. Ishihara, T. and S. Yamamoto, “Optimization of Mono-clonal Antibody Purification by Ion-Exchange Chroma-tography - Application of Simple Methods with LinearGradient Elution Experimental Data, J. Chromatogr. A,1069, 2005, pp. 99-106.

17. Yamamoto, S. and A. Kita, “Theoretical Background ofShort Chromatographic Layers. Optimization of Gradi-ent Elution in Short Columns,” J. Chromatogr. A, inprinting.

18. Guidance for Industry PAT - A Framework for InnovativePharmaceutical Development, Manufacturing, and Qual-ity Assurance, U.S. Food and Drug Administration, Sep-tember 2004.

About the AuthorsSei Murakami, PhD, is a General Managerof Industrial Systems Div., Hitachi, Ltd.,Tokyo, Japan. He earned a BS from OsakaUniversity, Japan, and holds a PhD in engi-neering from the Yamaguchi University, Ja-pan. His responsibilities include pharma-ceutical manufacturing plant development,engineering, design, and construction. He

participated in a mammalian cell culture research project atMIT in 1989. He is a member of ISPE, PDA, ASME, JapaneseAssociation for Animal Cell Technology, Institution of Profes-sional Engineers, Japan. He can be contacted by email:[email protected]

Hitachi, Ltd., 2-9-7 Ikenohata, Taito-ku, Tokyo, 110-0008Japan.

Peter Watler, PhD, is a Senior Director ofManufacturing, VaxGen Inc., South SanFrancisco. He earned a BS from the Univer-sity of Toronto, an MS from the University ofToronto, and a PhD in engineering from theYamaguchi University, Japan. He can becontacted by email: PWatler@ vaxgen.com

VaxGen, Inc., 347 Oyster Point Blvd., Suite102, South San Francisco, California 94080 USA.

Takashi Ishihara, PhD, is a Research Sci-entist of Pharmaceutical Division, KirinBrewery, Takasaki, Gunma, Japan. Heearned a BS from Kyusyu University, Japan,an MS from Kyusyu University, and a PhD inengineering from the Yamaguchi Univer-sity, Japan. His responsibilities includebiopharmaceutical purification process de-

velopment, and he has accomplished various process develop-ment and manufacturing of biologics. He is a member of theSociety of Chemical Engineers, Japan. He can be contacted byemail: [email protected]

Kirin Brewery, Co., Ltd., 100-1 Hagiwara,Takasaki,Gunma, 370-0013 Japan.

Shuichi Yamamoto, PhD, is a Professor ofBiochemical Engineering, Department ofChemical Engineering, Yamaguchi Univer-sity, Ube, Yamaguchi, Japan. He earned aBS from Kyoto University, Japan, and an MSand PhD from Kyoto University, Japan. Hisresearch interests include engineering analy-sis of chromatography and drying of biologi-

cal products. He is a Director of the Society of ChemicalEngineers, Japan, an official of the Japan Society for FoodEngineering, and a member of AIChE. He can be contacted byemail: [email protected].

Yamaguchi University, Tokiwadai, Ube, Yamaguchi, 755-8611 Japan.

Technology Assessment Diagrams


Project Scope Determination and CostControl Using Technology AssessmentDiagrams – A BiopharmaceuticalExampleby Mary Ellen Craft

This articledescribes theuse of atechnologyassessmentdiagram in thebiologics andpharmaceuticalindustry, andpresents thesteps todevelop, andevaluate thescope optionsfor a facility orprocess.

Introduction

How does your company make decisions?Can you locate the decision history ofyour company? Are those decisionsshared with the team?

“One would think that employees had acommon process to sort, organize, andanalyze information. Think again. Morethan four-fifths of both managers andworkers, when asked if such a practiceexisted in their organization, either said“no” or reported that they did not know.Of those who stated such a practice didexist in their organization, 31% of work-ers and 29% of managers said their man-agement takes no action to ensure that theprocess is used, or, if it does, they don’tknow about it.”1

Every team can improve its performance byutilizing simple communication and controltools. A technology assessment diagram is oneof these tools. It provides a one-page overviewthat allows the team to see, analyze, and makedecisions on almost any topic.

Ron Evans, a partner at Kepner-Tregoe®,which is a management consulting and strat-egy company, in an interview in Executive Fo-rum, stated:

“The use of critical thinking to addressbusiness issues seems to be lacking. Wefind that businesses are very good at try-ing to understand the content around anissue, such as the facts and technicaldata, but they are sometimes deficient in

the use of a common process to organizeand analyze that information.” 2

Each person thinks differently from another.Each person uses his or her own methods tomake a decision. Team decisions do not justhappen. In order to allow a team to jointly makedecisions and remain unified in those deci-sions, it is a valuable, almost imperative asset,to possess a common method for using a visualtool. This tool should aid in the decision-mak-ing process, record the options considered, andidentify the solution(s) determined. A visualtool also may be needed when communicatingdecisions/choices, and the justifications for thosedecisions to management.

Management is more often demanding, “Inot only want to see what the decision is andwhat information you’ve got, but I want to seehow you processed that information in yourrecommendation.”2

This article presents a modified decisionmatrix with the use of a weighted rankingsystem to determine and analyze the targetscope/functionality or user requirements ofmany desired components or parts necessaryfor the construction of biopharmaceutical fa-cilities.

Many decision-support software programsare available on the market. Most of theseprograms are based on a decision matrix or adecision tree. However, “Computer programscannot set goals, think up alternatives, assignweights, or evaluate criteria. You have to dothat.”3 Computer programs can be valuable incrunching numbers and asking questions thatmight be overlooked without the program. How-





ever, decision support software only allows one to see andwork with one column or one row of the decision matrix at atime. While discussing the benefits and drawbacks of com-puter decision-support programs, Jared and William Taylor,as authors and management consultants, further state that:

“In the final analysis, all a decision matrix does is sumup weighted attributes of the alternatives. Why can’t youdo this with a spreadsheet? You can…spreadsheets canbe made to function as a decision matrix.”3

This article presents technology assessment diagrams and adecision-making process that can easily be used by yourteam. This simple tool utilizes a spreadsheet as a decisionmatrix.

What can a Technology AssessmentDiagram Do for Me?

A Technology Assessment Diagram (TAD) can help you makedecisions, capture your project requirements or components,and document solutions. It also can help you to keep yourproject within budget. A TAD gives your team members,whether internal or external, a visual way to consider alter-natives or options and view, at a glance, the cost and technol-ogy impact of those alternatives.

TADs aid in each of the following:

• communication within team and to management• defining and documenting physical and functional options• determining target functionality or baseline scope• analyzing scope content, its quality and associated costs• controlling project costs

• providing support documentation for the following:- project estimates- change management systems- value engineering exercises

After a project is defined, any team can use TADs for furtherdefining, analyzing, controlling, and tracking components ofthat project.

What is aTechnology Assessment Diagram?

A TAD uses a spreadsheet as a decision matrix. It uses astructured methodology to assist in decision making whileproviding a documented trail of scope options and theirimpact on project cost. See Figure 1 for layout of a basic TADtemplate.

How do I Construct and Use aTechnology Assessment Diagram?

The steps in developing TADs are:

1. Begin with your project specific template2. Divide your project into parts to be analyzed and begin a

TAD for each of these parts3. Identify attributes for each part, area, system, or function

- Assign weights to the attributes4. Further define the scope by developing scope function

options- Select target functionality (baseline scope)

5. Evaluate and analyze your options- Cost Each Option- Apply Ranking

Figure 1. Technology Assessment Diagram (TAD) basic template.



- Analyze Options by Attributes6. Justify or alter the scope target functionality

This article uses the design of a new biopharmaceuticalfacility as the basis for examples given. This article illus-trates how to use TADs. Your team can use them to establishthe quality level of construction materials, systems, andequipment as well as to determine and document the func-tional and aesthetic aspects of project components.

1. Begin with your Project Specific TemplateInsert your project specific information in the header andfooter with information such as your company name and logo,project name and logo, project number, location and/or build-ing name and number, page number, key to grading scale,author’s name, date generated, revision date or log, and filename and path. Refer to Figures 1 and 2, Item A: Header andItem B: Footer.

2. Divide your Project into PartsAfter your template is set up, begin dividing the project intoareas, divisions, categories, or items to be analyzed. Next,identify the area or item being evaluated. For a facilitydesign, a TAD can be used to evaluate the physical parts andsystems for your proposed facility. The planned facility couldbe divided into physical parts such as structural systemsexterior building walls, roofing, doors and windows, case-

work, landscaping, signage, and interior materials, and roomfinishes by area, robotics, processing equipment, or coolingtowers. It also can be divided into building systems, such asautomation, unit operation integration, power, emergencysystems, water and other utility systems, chemical storage,chemical distribution, waste treatment, and security sys-tems. Another division could be functions such as sampling,warehousing, maintenance, or training. In other words, yourimagination and your needs are the only limits to its applica-tion. Examples of possible areas and functions are shown inTable A.

Enter both the area and the function to be analyzed. Referto Item C in Figures 1 and 2. After the main parts, areas, andprocesses are identified, construct a TAD for each.

3. Identify Attributes for each Part, Area,System, or FunctionMake a list of attributes that apply to each of your projectdivisions, categories, or sections to be analyzed. Examples ofattributes are:

• Appearance • Image• Chemical Resistance • Life Expectancy• Cleanable • Maintainability• Code Requirement • Match Existing• Company Standard • Reliability• Constructability • Security

Figure 2. Project specific TAD template.



• cGMP Requirement • Service Support• Delivery Time • Size• Durability • State of the Art• Ergonomics • Spare Parts• Expandibility • Training Need• Flexibility • Upgradable

Note: Attributes can and will differ from physical part tophysical part, system to system, or function to function.

For each specific project part, area, system, or functionbeing diagrammed, enter the most important attributes ascolumn headings. Refer to Item D in Figures 1 and 2. Eachcolumn becomes an attribute to be evaluated. Cost is alwaysthe final attribute and usually the most important. Refer toItem E in Figures 1 and 2. Operating, replacement andcapital costs can each be a separate cost attribute. Note:Attributes can be whatever is important to the team for eachspecific project component or category.

After attributes have been assigned, they should beweighted with more points assigned to the more importantattributes, in preparation for analyzing the options later.Note: If your team decides to use Kepner-Tregoe® Matrix or K-T Analysis,4 then there is an additional step at this point todetermine which attributes are essential (must haves) andwhich attributes are desired, but not essential (wants). Kepner-Tregoe® (K-T) Analysis is a methodology for identifying andranking factors critical to a decision. See Table B for the basissteps in K-T Analysis.

4. Further Define your Scope by DevelopingScope Function OptionsA TAD is a spreadsheet which charts functional or physicaloptions. These options are identified and recorded on theassessment diagram. This data is gathered through inter-viewing users, holding group brainstorming meetings, refer-encing documents such as your company standards or plan-ning documents such as P&ID’s, or from viewing the area, thesystem, or activity in actual operation.

Brainstorm options, then list the different options in thefar left column. Refer to Item F in Figures 1 and 2. Optionsmay then be sorted, usually with the most extravagantsolution at the top and the most utilitarian option at the

bottom. Then number the options for ease of discussion.Select target functionality. This will be the option that is

either covered in the budget, the company standard, or theleast (lowest) option that appears to meet the project and userrequirements. Note: The team may need to evaluate alloptions before determining the final target functionality.

5. Evaluate and Analyze your OptionsCost options - before analyzing the options as a whole lineitem, price each one. Determine the total cost of an option ifpossible. If the options are not yet designed, then price a unitof that option. If total capacity, quantity, size, or area isunknown, then use unit prices that can be easily compared.For example, for physical items, one could price units such asper Square Foot (SF) or per Linear Foot (LF), per ton or perone lot of a determined number of items, such as 100 pieces.

Select and implement a ranking system. Either subjectiveor objective scales may be used. Whether your team uses asubjective or an objective scale, care should be taken to use aneven number of rankings so that the team must determine ifsomething is above or below average. An odd number ofrankings allows the team to choose the middle ranking andthereby “sit on the fence.” A subjective scale uses rankingssuch as low, medium, high and highest or poor, fair, good, andexcellent. Note: The grading scale for different attributesmay vary in a subjective ranking system. Also, subjectiverankings may be reversed on different attributes. A lowranking may be good for one attribute and bad for another. Itis easier to score a subjective scale, if numbers are applied tothe ranking. In other words, low could equal 1 (or 10),medium could equal 4 (or 7), high could equal 7 (or 4), andhighest could equal 10 (or 1). So, if you have a reverse scale,be careful that numbers have been properly assigned beforeadding the total score.

It is difficult to remain objective, especially when a deci-sion impacts you, your team, your facility, and your company.One way to take some of the subjectivity out of the equationis to use an objective ranking system. An objective scale usesnumbers for ranking. Just remember to use an even numberof numbers. An even number of choices requires the rankingto be more one way, either better or worse, than the other. Inother words, you cannot select the middle number, because it

Site Areas Building Areas Systems Equipment

Roads and Parking Manufacturing Purified Water Fermenter

Central Utility Laboratory HVAC BoilerBuilding

Administration Administration Lighting Cooling TowerBuilding

Production Packaging Fermentation CIP SkidBuilding

Shop and Receiving Communications AHU’sWarehouse

Building G Mechanical Security MCC’sSupport

Table A. Examples of areas and functions (Item C).

K-T Analysis Technology AssessmentDiagram

1. State the Purpose Separate Project Scope into Parts

2. Establish Objectives Establish Requirements orAttributes

3. Classify Objectives by MUSTs Explore Optionsand WANTs

4. Weigh the WANTs Assign ranking Weights to theAttributes

5. Compare Alternatives Analyze Options with WeightedScores

6. Choose the Best Course of Action Determine Target Functionality

Table B. Comparison of steps in K-T Analysis vs. a TAD.



isn’t there. One to 10 is the most common scale used. Thisauthor recommends the use of an objective ranking system,such as K-T Analysis.4,5 A comparison of the step descriptionsfor K-T Analysis and the steps in utilizing a TAD are shownin Table B.

In order to perform an evaluation with K-T Analysis, onemust first identify the items that are must haves. Refer to Step4 in Table B. The must haves are absolute items that arerequired no matter what. Therefore, the option will either meetthe criteria or not. If the option meets the must have attribute,it is a yes and remains in consideration. If the option does notmeet the must have requirement, then it is rejected and nolonger considered. If an attribute is not a must have, then thatattribute becomes a want, or a desired but not necessary item.(Refer to Step 5 in Table B.) The want attributes are thenranked on a scale of 1 to 10, with 1 low and 10 high. Therankings are multiplied with the weighted score of eachattribute and the scores are tallied. The highest score wins.

Information blocks on the TAD can be analyzed with theK-T Analysis or another objective ranking system. The evalu-ation can be more objective using attributes where actualnumbers can be applied, such as GPM, cost data, ROI, yearsof life expectancy, light reflectance, or assigned points for agiven range, etc.

Assign a ranking (1-10) to each block where an optionintersects an attribute. Note in our completed TAD exampleshown in Figure 2, not only is the ranking shown, but anexplanation is written in each block. The explanation canhelp the team remember why and how the rankings wereapplied and identify the justification for those choices.

Analyze the evaluated options by multiplying the optionratings with the attribute weights. Enter the score for eachscope function option. Using a TAD, the team can immedi-ately see the benefits, restraints and cost impact of options foreach component that was analyzed.

6. Justify or Alter the Scope TargetFunctionalityYour target functionality option on each individual TADshould coincide with the option gaining the highest score. Ifthey are not the same, then check all items on both theoriginal target function option and the final preferred optionwith the highest score. Verify that you have selected the leastscope that will meet your specific project requirements. If thehighest score is the least scope that meets project require-ments, then change your target functionality to that option.However, take time to justify this change. Take extra care inthis step to ensure that your subjective self does not overruleyour objective analysis.

Benefits realized from the use of TADs are:

• Communication – as a tool, it is one of the best ways toconvey design requirements to the owner/user and toconvey his/her needs and choices to the design team -especially when used in conjunction with graphics,sketches, manufacturer’s specifications and catalog cuts,photographs, samples, or color and material boards.

• The full team can easily view the same data in a logicalformat

• Documented basis of the team decisions – the TAD servesas a reminder to the design team of the project require-ments, including quality and cost of the components.

• Documentation of target scope/functionality and devia-tion from that scope – the TAD traces the decision-makingprocess and makes the team justify deviations from theoriginal target functionality.

• Basis of review and control of changes during constructionand in future alterations can be used as a check of theconstruction estimate material content. The TAD docu-ment not only can help track scope changes, but also canshow the impact of those changes on the project cost. Oneof the ugliest enemies of any project is scope creep thatescalates the cost of the project. Scope creep can be causedby increasing size or quantity, by raising the level ofquality, customizing the solution, or by chasing that illu-sive special image.

• Aid to cost control of project elements.

• TADs can be used later in value engineering exercises.They can remind the team of options that have alreadybeen evaluated and hopefully save time by keeping theteam from repeating decisions that have already beendetermined.

The major benefits derived from this method are scopeclarification, efficient programming, promotion of teamwork,and gaining the buy-in of all stakeholders. An added benefitis that the graphics are great for presentations, which areoften necessary to secure financial appropriations.

SummaryIn summary, the information contained in your TAD shouldgive a clear understanding of what is included for componentsof the project scope. This decision-making method removesmuch of the subjectivity and replaces it with objective reason-ing. An important team goal to remember is to design to meetrequirements and remain within the budget. TADs help dothis and more. They serve as a graphical tool to track thedecision-making process and to control material, equipmentand system design, automation, or other quality upgrades.

This tool, especially when used in conjunction with graph-ics, sketches, catalog cuts, photographs or color and materialboards, is one of the best ways to convey design requirementsto the owner/user, and to enable them to convey their needsto the design team. Size or quantity is usually controlled withreview of plan view drawings, elevations, and building sec-tions. Layouts give square footage of areas and relativeequipment size and their quantity, which are generally usedas a basis for cost analysis of the project. However, floor plansdo not indicate the quality of materials and equipment. A



change in equipment model number, software, custom de-sign, material, or finish can greatly affect the overall cost ofthe project. TADs are useful for defining scope and targetfunctionality, communicating the impact of change, facilitat-ing and documenting group decisions, and for controllingproject scope and costs.

References1. Middlebrook, John and Tobia, Peter, “Decision-Making in

the Digital Age,” USA Today, September 2001, Volume130, Issue 2676, p. 50.

2. Evans, Ron, Executive Forum interview, “Critical Think-ing: Putting Your Heads Together,” Management Review,November 1997, Volume 86, Number 10, p. S1(3) Bis. Coll:105U1721.

3. Taylor, Jared and Taylor, William, “Searching for Solu-tions: Decision Support Programs Can Give You Answers.They Can Also Prevent Risks,” PC Magazine, September1987, Volume 6, Number 15, p. 311 (27).

4. Kepner, Charles H. and Tregoe, Benjamin B., “The NewRational Manager,” Princeton Research Press, Princeton,New Jersey, (1997).

5. “Steps to Approach Decision Analysis with Kepner-Tregoe®,” www.valuebasedmanagement.net.

About the AuthorMary Ellen Craft received both a Bachelorof Architecture and a Bachelor of AppliedArts, which is a shared topic major in interiordesign, art, and architecture, from the Uni-versity of Kentucky. She is currently a ProjectDirector in Fluor’s Life Sciences Group. Sheis experienced in design management forbiotech and pharmaceutical facilities in the

U.S., Ireland, and Spain. Craft has developed and deliveredmultiple seminars on project management and technicaltraining for the pharmaceutical industry. She has been acourse leader and speaker for ISPE at the Chapter andinternational level. Craft has also spoken at managementmeetings for several pharmaceutical manufacturers. Sheformerly served as president of the ISPE Great Lakes Chap-ter, and currently chairs ISPE’s Editorial Committee andPharmaceutical Engineering Subcommittee. She is also amember of the Pilot Journal Task Team and lead author ofthe regulatory chapter in the Laboratory Baseline® Guidenow under FDA review. Craft is a Registered Architect, anda certified Project Management Professional (PMP). Shemay be contacted by telephone at 1-864/281-4605.

Fluor Enterprises, Inc., 100 Fluor Daniel Dr., MS-C202D,Greenville, South Carolina 29607 USA.

Improving Documentation Structure


Practically Validated – Maintainingthe Tested Baselineby David Paspa

This articlepresents waysto improvedocumentationstructureincorporatingrequirementstraceability andrisk analysis.

This article offers some simple and spe-cific ways to improve documentationstructure and incorporate requirementstraceability and risk analysis. It pro-

vides tools that can be used to improve the levelof compliance to perform the job correctly fromthe beginning.

The techniques defined can be used for anytype of validation project including computervalidation, equipment qualification, and pro-cess validation.

For many, validation is still a grey area andthere are good reasons for this. Pharmaceuti-cal and biotechnology are diverse industriescomprising:

• many different and sometimes complex pro-cesses

• most engineering and scientific disciplines• sophisticated regulations that govern the

manufacture of products which are scat-tered throughout a variety of sources andoften require a fair amount of interpretation

It is probably impossible to find a single personwho truly understands all of the chemistry,engineering, and governing regulations relatedto drug manufacturing. To complicate this fur-ther, manufacturing companies often sell tomultiple international markets where regula-tory expectations and enforcement varies.

ValidationThe term validation itself is poorly under-stood at best. One classical definition is:

“…to provide documented evidence whichprovides a high degree of assurance thatsystems, operated within their specifieddesign parameters, are capable of repeat-edly and reliably producing a finishedproduct of the required quality.”

That’s all well and good, but what should peopleactually do? Unfortunately, the term and itsdefinition do not give specific direction on thephysical tasks to be performed. Perhaps for thisreason, it is often left until later in the projectplanning process. Sometimes it is just an after-thought, applied for the wrong reason – be-cause of fear of the regulators rather than as ameans of self-assurance.

QualificationMany of the same comments that have beenmade about validation also can be made aboutqualification. While the term is a little vague,the physical activity that people actually do istesting.

What’s the difference between qualificationand testing? Qualification involves testing sys-tems to demonstrate they do what they aresupposed to. In other words, Qualification is

Figure 1. Validation =Testing + Management.





testing.Testing has meaning only when systems are tested against

what is required of them. It must first be stipulated “this iswhat it is supposed to be” and then tested to show “this is whatit is.” There is no point testing to say “it is what it is, whichmust be what it is supposed to be” and yet this approach is stillused frequently.

An example would be to say “Test that chair.” It would bedifficult to make a sensible test because there is uncertaintyas to what to look for. However, if it was first specified that“This chair is designed to support the weight of an 80 kilo-gram/176 pound person,” then it would now be possible todevise a rational and quantifiable test that can measurewhether the design intent has been accomplished.

So, a qualified system is simply a tested system.

Qualification = TestingIn order to test anything, in any industry or context, therequirements must be defined first. This is a fundamentaland important point. It is not meaningful to test somethingunless it has a specified requirement.

Why Test?When a system is tested a tested baseline is achieved. Fora given set of inputs, the system has a predictable responseand provides a known output. Any test result for that systemis valid over time provided the system does not change.

Once a system has changed, the test may or may not bevalid. A judgement based on the nature of the change wouldneed to be made to determine whether the test results werestill considered valid or whether the system would need to bere-tested to find out if that same result is received the secondtime around. The change may be such that a new test needsto be devised to demonstrate some new system requirementsor attributes.

It requires the investment of significant time and moneyto achieve a tested baseline through a rigorous program ofspecification and testing. Therefore, it makes sense to protectthat asset by managing the system so there is confidencethat the tested baseline is current over time.

In order to achieve this, all aspects of the system need tobe controlled, including:

• the physical components of the system• the people who use and maintain the system• associated information and documents• ongoing changes made to the system, both planned and

unplanned

To summarize, an activity-based definition of validationconsists of testing and management to maintain the testedbaseline.

Testing and management are equally important - Figure1. A tested baseline that is not managed quickly becomesoutdated. Procedures that are implemented to manage asystem that hasn’t been properly tested, do not improve theassurance of the system response for a given set of inputs,

regardless of management efforts.The tested baseline should be thought of as a physical

thing, such as a ball. Don’t drop it! The majority of thediscussion that follows proposes ideas and techniques thatcan be employed to develop and maintain the tested baseline.The format of the tested baseline and the way in which it iscreated are critical factors in its ongoing maintainability. Theideas presented are intended to promote and facilitate thismaintainability.

Most “validation” projects are in fact “qualification”projects. There is often very little management of the testedbaseline that is handed over at the end of the project. As aresult, the tested baseline is nearly always compromised withthe passing of time resulting in systems “falling out” ofvalidation. This usually results in the whole qualificationexercise having to be repeated.

To avoid this situation, it might be useful to focus on theactivities being performed. Rather than describing a systemas “validated,” as if it were a property of the system, it wouldbe better practice to say the system is “under validation.” Thisbetter indicates there is a method in place to continuouslymanage and control the system in an ongoing way to keep thetested baseline current.

It is interesting to note that testing and management arecommonly understood activities which have been performedby humans for thousands of years to achieve some quiteremarkable things. When good science and engineering andgood project management are used, validation is nothing newand nothing extra.

Now, in order to formulate meaningful tests, there mustbe pre-determined requirements. There must be a specifica-tion that says “this is what it is supposed to be” and then acorresponding test that shows “this is what it is.”

• If there are no specified requirements, there can’t bemeaningful testing.

• If there are no meaningful tests, it is not possible toachieve a tested baseline.

• If there is no tested baseline, there is nothing to manage.• If there is nothing to manage, the system can’t be “under

validation,” i.e., under control.

Therefore, it can be deduced that requirements are funda-mental to validation. And yet, it is still common to find“validated” systems with no definition of what the system issupposed to do.

RequirementsWhen writing requirement specifications, it is crucial toremember the document is not the job. The purpose of theexercise is in fact not to write a document, but to conveyinformation to the reader of that document so they under-stand what is required. The document exists to be read, notwritten.

To facilitate understanding, it is good practice to:

• use simple short statements



• keep each premise separate• stick to the facts; less text gives rise to more understand-

ing

One approach to writing specifications is to number indi-vidual requirements to aid discussion and traceability. Thisis the same approach used in numbering equipment, likevalves and pumps, with a unique tag number - Table A.

Unique, clearly identified, and testable specifications pro-vide greater understanding to all. There is no point in havinga specification or requirement if it cannot be tested in someway. How can it be confirmed as even having been deliveredby the supplier? In fact this does not just apply to validationin the pharmaceutical industry. Regardless of the industry,at some point suppliers expect to be paid for goods deliveredor services rendered. It is common sense to assure oneselfthat the product is what was wanted and is what it purportsto be before it is paid for. This is just prudent contract andfinancial management.

One way to ensure requirement numbers are unique inthis way is to use a dynamic outline numbering field code togenerate the serial number. These are available in most wordprocessing programs. When the document is complete, thesedynamic field codes can be unlinked which converts them tostatic text. From that point on, each reference number is justplain text and is inexorably linked to the correspondingrequirement text. The reference number can then be safelyused to refer to a requirement from outside the documentwith confidence that the reference cannot be broken.

New requirements can still be added if there is a change tothe document and they take on the next highest unused serialnumber. Also, old requirements and their reference numberscan be deleted. Thus, the requirement reference numbersmust be unique, but they do not have to be consecutive. Also,numbers may be missing if superseded requirements havebeen deleted. Therefore, to be able to find a specific numberand therefore a specific requirement, a Requirement Refer-ence Number Table of Contents is used to list all of thenumbers in order and bookmark its page - Table B.

V-ModelThe V-Model defined in the Good Automated ManufacturingPractice (GAMP® 4) Guide provides a structured frameworkfor specifying requirements and then testing them to demon-strate they have been correctly delivered. In its simplestform, it defines the phases shown in Figure 2.

However, design is a continuous process. At best, the V-Model is a quantum approximation of what is actually goingon during the design continuum. While requirements are“lumped” into User Requirements Specification (URS), Func-tional Specification (FS), and Design Specification (DS) docu-ments and drawings, the real-life process is actually twosteps forward and one step back (often with a side step thrownin for good measure). Even agreeing and documenting theuser’s requirements in the URS can be a difficult task whenthere are a number of stakeholders involved.

Keeping that in mind, how does the V-Model assist in the

design process? The advantages are:

• It clearly separates requirements which have a differentpurpose.

• It clearly separates requirements which have a differentimportance.

User RequirementsNot all of the design information for a project is of the sameimportance. The user requirements are the most importantbecause they define what the process requirements are. The“process” may be a production process to manufacture apharmaceutical product or may be a business process such aschange control. Regardless of what gets implemented, it isthis “process” that must be delivered by the designer. Theserequirements are fundamental to the user’s business and arenot-negotiable as far as the designer is concerned.

Any parameters which are critical to achieving the user’sdesired level of GMP should be clearly identified in the URS.

Functional RequirementsThe second level of importance is the functional require-ments. These define what has to be done to implement theprocess, but at this point don’t stipulate how to do it. This isalso referred to as conceptual design.

These operational requirements are often important tothe client, but less so than the user requirements. Forinstance, if another concept which saves time and money wasproposed, the client may accept the alternative to what theymay have stipulated they initially wanted because it is good

Ref Requirement

U12 Purified Water shall have a microbial load of £ 0.1 cfu/mL.

Table A. Requirement reference numbers.

U1 ............................................................................................................. 7U2 ............................................................................................................. 7U3 ............................................................................................................. 7U4 ............................................................................................................. 8U5 ............................................................................................................. 8

Table B. Requirement reference number table of contents.

Figure 2. GAMP V-Model.



practice to save time and money.Note the functional requirements define what has to be

done, but not how to do it. There may be multiple designsolutions to implement the operations required, but these arenot determined at this level. Try to focus only on what has tobe achieved.

The FS should comprise unique, clearly identified, andnumbered requirements similar to the URS. In addition, thefunctional requirements should be traceable to the higherlevel user requirements. This parent/child relationship orhierarchy is inherent in the V-Model from URS to FS to DS -Table C.

Notice in the example above that the cross-reference to theparent specification requirement has been placed right intothe child specification document. This is a useful tool that canbe used to ensure all parent requirements are addressed inlower level design documents.

Design RequirementsThe design requirements are the third and lowest importancerequirements. For the first time, a definition is providedspecifying how the operations identified in the FunctionalRequirements are going to be implemented. The designrequirements are the most negotiable and offer great oppor-tunities for time and cost saving.

The same format and cross-referencing methodology canbe used to define these requirements which are referred to asdetailed design - Table D.

Design requirements talk about how the functional re-quirements will be implemented using physical objects. Theydefine what objects will be used and specify their configura-tion and orientation.

Software Design SpecificationFor a software project, the DS should talk about softwarestructure, software modularity, data definition, and dataflow. In this case, the design specification is especially crucialbecause it is the only physical embodiment of the structure of

Table C. Functional requirement reference numbers and traceability.

Parent Ref Requirement

U12 F201 The system shall maintain a line velocity of 1 m/s in allpipework.

Table D. Design requirement reference numbers and traceability.


F201 D177 The supply pump shall be rated to deliver 120l/min at 5bar.

Figure 3. Design development.

the code. A mechanical design has piping, pumps, and valves,and it can be physically seen how the design has beenmanifested in the real world. In a software project, there isnothing to “see,” and hence, the design specification is all themore crucial.

Prototyping and R&DSometimes it is essential to prototype or pilot the designbefore starting to write the DS. Prototyping is an excellentdesign tool, but problems can arise when considering thewasted time and effort writing down things that may never beused. Does the prototype design have to be documented? It ishighly likely the design will change dramatically as a resultof the prototyping exercise.

To address this issue, consider what is involved in specify-ing anything. Key elements are education, research, consul-tation, experience, assumptions, etc. The reality is that all ofthese things are the sum total of people’s trial and errorexperiments over time. A vast amount of what is known todaywas found out by just trying. Thus, R&D and prototyping arein fact usual and valid activities to help define a specification(i.e., a bit more trial and error). Design does not just includewriting a document in a word processor. All kinds of activitiesand types of information can be described as “design,” includ-ing drawings, calculations, models, lists, data, etc.

The amount of up front documentation without R&D effortdepends upon the certainty and amount of knowledge thedesigner has of the final design. If nothing is known, it isprobably a good idea to spend some effort playing around withcomponents to learn how they behave before starting a formaldocumentation process. If a lot is known about the system,usually because the designer has done something similarbefore, then the documentation process can be started at thevery beginning with a high level of confidence that the workwill be useful.

For process applications, R&D from pilot plant or labora-tory batches may be the only way to identify and characterizethe critical process parameters. The process design shouldtake into account these critical parameters and mitigatetheir risk to the product or process through a well thought outrisk mitigation strategy. The knowledge gained from thisR&D exercise allows resources to be directed at managingand testing these critical points in accordance with a sensible,risk-based approach.

While it is good practice in any R&D effort to record whatwas done to retain and reuse the knowledge gained from theexperiments, it is not practical to try to formally specify upfront all aspects of the prototype design because they are justnot known.

Separating InformationMany organizations cannot or do not distinguish between thediffering levels of requirements and priorities. It has alreadybeen established that when trying to convey what is impor-tant to the reader it is important to keep things short andseparate. Therefore, it is of the greatest advantage to keep thedesign requirements (the most negotiable requirements)



separate from the process requirements (which are not nego-tiable). Simply by calling these varying requirements differ-ent things conveys important information.

The amount of information put into the URS depends onthe project execution model and the purpose the URS isintended for.

• Issue the URS to a supplier – in this case, it is importantthat the supplier is told enough to ensure clients get whatthey want. This would comprise the process requirementsand also identify any functional and design constraintsthat the supplier must work within; i.e., the client is notgiving the supplier a blank sheet to start from. Sometimesa client can have a supplier do the work of writing the URSon their behalf, but the client must still approve thedocument and take responsibility for its content.

• Use the URS as part of client’s in-house project – in thiscase, just include the process requirements. Any func-tional constraints and design constraints can be directlyincluded in the FS and DS documents that the clientcontrols.

Whatever the approach, it is important to note that informa-tion can be conveyed to the reader simply by the structure ofthe documentation.

Document ModularityIt is advantageous to keep documents modular and usemultiple “buckets” when following the V-Model. Using mul-tiple buckets also minimizes the scope of a change. Forexample, if there was a detailed design change, the designspecification would need to be updated, but the user require-ment specification and potentially the functional specifica-tion documents would not need to be changed. In this examplethe IQ on the affected component would presumably need tobe updated, but the OQ and PQ may be unaffected.

Time DependencyAn interesting question to pose is “Where is the time axis onthe V-Model?” It isn’t explicitly drawn on the diagram, but themodel does infer precedence from left to right. There isprecedence of document approval down the left hand side ofthe V and precedence of order of testing IQ, OQ, to PQ goingup the right hand side. This precedence must be adhered tobecause it is a regulatory requirement.

A fast track project demands an early start on all activi-ties, design and construction included. However, an idealquality-based project, would demand a late start to ensure a

succeeding phase wasn’t started unless the preceding phasewas reviewed and approved. The reality is that most projectsdemand a compromise between these two scenarios to man-age what is considered an acceptable level of business risk.Similarly, nowhere does it say that succeeding documents inthe V-Model can’t be started before preceding documents arecomplete, but the level of business risk needs to be managed.However, the precedence of document approval and testingmust be adhered to.

Legacy SystemsThese are really the ultimate in project fast tracking. Nosooner are the documents started than the system is alreadybuilt; i.e., time is effectively compressed to zero. So how doesthe V-Model apply to a legacy system? The exact samestrategy and documentation approach can be implementedfor a legacy system as for a prospective new system. In thiscontext, a legacy system is considered to be any system whichalready exists.

In these cases, it may be possible to justify a reducedamount of testing as part of the risk analysis if pertinenthistorical data is available for normal operation of the systemallowing a focus on the system’s abnormal condition handlingto demonstrate robustness and reliability. Legacy systemqualification offers an ideal opportunity where a sensible,risk-based approach can dramatically reduce the workload.

Risk AssessmentThe unique reference numbers for each requirement cancontinue to be used as part of the risk assessment process.This risk analysis is usually performed during the conceptualdesign of the project to “design out” high-risk situations in thefirst place. The designer’s goal is for the true quality assur-ance and validation of the system to be intrinsic in thesystem’s design, rather than being in some associated docu-mentation on a bookshelf.

The results of the risk assessment also can be used toidentify those requirements which will not be formally testeddue to a determination of risk to the process or system. Thereare many methods available to perform risk analysis, such asaccording to a calculated risk priority as defined in GAMP 4.

Typically, requirements can be grouped according to sys-tem parameters and their failure modes. This can be used aspart of an argument to reduce the scope of abnormal conditiontesting by justifying why certain test cases do not need to beexecuted. For instance, it may be stipulated that only HighRisk Priority failure modes need to be tested and these mightrelate to parameters with a High Risk Classification and aLow Probability of Detection. GAMP suggests the use of a

Ref Parameter Risk Justification Failure Mode Effect Probability of Risk PriorityClassification Detection

F371 Purified Water High Direct indication of Input C028 Standard PLC function High LowF372 conductivity current water under-range or detects bad input andF373 directly after the quality over-range raises alarm. PW noF219 RO/EDI plant. longer available to users.

Table E. Risk assessment.



medium level risk priority as well.There are many risk assessment approaches available,

also commonly looking at impact of a failure, likelihood offailure, frequency of occurrence, residual risk after correctivemeasures are put in place, HAZOP, HACCP, etc. The Internetis an excellent source of information. At the end of the day, asimple approach is often the most sound.

Requirements TraceabilityRequirements traceability refers to the process of trackinghigher level requirements down through the lower levelrequirement documents and even across to the associatedtest documents.

For instance, why is a certain pump put in a certainposition? Or why has a certain piece of software been pur-chased or developed? Lower level design elements are in-cluded because of a higher level functional requirement.Similarly, functional requirements come about because ofhigher level user requirements (i.e., process requirements).It is prudent to check that all of the higher level requirementshave been addressed by lower level conceptual and detaileddesign elements and make sure that nothing has been over-looked.

Requirements traceability is usually addressed by cross-referencing the section heading numbers in two or moredocuments in a table or matrix. Unfortunately, this is aretrospective method. Also, using section headings is peril-ous because when someone updates the document, they canforget that inserting a new section will automatically renum-ber the existing sections in the document and place theRequirements Traceability Matrix (RTM) out of date.

A prospective method of requirements traceability can beused as a useful tool to ensure that all requirements havebeen addressed as the documents are being written, not afterthey are complete. The unique reference number applied toeach requirement in the specification, as shown previously,can be used to provide the vertical requirement traceabilitydown the left-hand side of the V-Model.

A functional specification might look something like thatshown in Table F. The document has two columns on the lefthand side in front of each written requirement; a parentrequirement reference number, and a unique child require-ment reference number. Requirement reference numbersmight be F1, F2, F3, F4, etc. up to, say, F203. Similarly, theuser requirement specification numbers might be U1, U2,U3, U4, etc. up to, say, U50. When writing the functionalspecification, the text U12 would actually be written againstthe corresponding functional requirement F201 to demon-strate the link between the higher level URS. Similarly, U13and U27 might correspond to F202. U10, U27, and U41might correspond to F203, etc. All requirements F1, F2, F3,etc. and U1, U2, U3, etc. can be cross-referenced in this way.For some reason, F200 may not directly relate to a parentrequirement.

The child requirement reference numbers in the documentmust be unique, but may cross-reference with multiple parentreference numbers. Thus the requirement traceability infor-mation from a higher level specification is put right into thelower level requirement specification as it is being written.This provides traceability down the left-hand side of the V.

Requirements Traceability MatrixTabulating the above information gives rise to an RTM. Thisprovides an overview of the traceability put into the require-ments specifications. This table can be quickly generatedfrom the requirements specifications many times while writ-ing the specifications to help close out the cross-referencing tomake sure all high level requirements have been addressedin some way.

Sometimes lower level functional or design requirementsare there for associated design reasons or for consequencesfrom the parent specification which have not been explicitlystated. In these cases, the lower level requirements do notmap onto a higher level parent requirement and a justifica-tion can be provided in the RTM as to why that is so. An


F200 The system shall do something.

U12 F201 The system shall maintain a line velocity of 1 m/s in allpipework to limit bacterial growth.

U13 F202 The system shall do this.

U27

U10 F203 The system shall do that.

U27

U41

Table F. Requirement cross-referencing.

UNRF JUST F200 F201 F202 F203

UNRF X

JUST A

U10 X

U11 X B

U12 X

U13 X

LEGEND: A: Functional requirement added due to existing operational constraint of target system.B: User requirement refers to project schedule and not an operational technical requirement.

Table G. Requirements traceability matrix.



excerpt of an RTM is shown in Table G.The RTM is useful evidence to contribute to the design

review process by demonstrating that all requirements havebeen addressed. Note that it only contributes to the reviewprocess because it doesn’t assure that the design sensibly orcompletely meets the requirements from a technical point ofview. Technical design reviews which test the design toensure it correctly implements the intended parent require-ments and complies with applicable statutory and GMPregulations still need to be performed.

Test ProtocolOnce the requirement specifications have been written andapproved in accordance with the appropriate design anddocument reviews, the issue of testing documents to demon-strate these requirements can be addressed.

Test protocols should contain the following informationfor each test:

1. A test method needs to be devised that will demon-strate that each requirement has been met. This methodshould be sufficiently detailed so that someone couldrepeat the test at a later date as it was first carried out.However, too much detail tends to become unnecessaryand even erroneous, causing test failures due to incorrecttest documentation. Over-specifying the test method alsocan lead to a blinkered approach by the tester duringtesting. Sometimes a compromise with some “monkey”testing is useful to have the tester use their initiative to tryto “break” the system in question by performing a series ofactions which could not have been foreseen and formu-lated when the test protocol was written.

2. An expected result should be stipulated which definesthe acceptance criteria for the test. This should bedetailed enough so that the tester can clearly identify apass or fail condition when the test is executed.

3. An area where the tester can record the actual resultshould be provided. These are the observations of whatactually happened during test execution.

4. There should then be a clear indication of the test result.

This should be an unambiguous statement of pass or failrather than a simple tick or other mark which could haveambiguous meaning.

5. There should be space for the tester to provide an initialand date for each test.

Table H shows the tabulated format.There should be a witness of the tests and the test

environment. The witness needs to supervise the testingprocess and needs to have final signoff on the results, espe-cially for critical tests. This verification by an independentwitness is performed in accordance with GMP regardingcritical steps in a manufacturing process.

The witness does not need to sign off on each and everysingle test, but can sign the protocol a page at a time or evenjust once on the front page for the overall document.

Test TraceabilityTest protocols can be structured in a similar way to require-ments specification documents with a unique test referencenumber that is linked to the parent document requirementnumber. Sometimes one requirement must be demonstratedover a number of tests or a single test may address multiplerequirements. An RTM can be generated in the usual way tosummarize these relationships.

The V-Model is usually drawn with dashed lines from thequalification steps back to the corresponding specificationsteps. Usage of the requirement reference number as a testreference number provides these links on the horizontal axisof the V-Model.

A design change can quickly be traced from the designspecification across to the corresponding tests using therequirement and matching test reference numbers. Redraft-ing and repeating the affected tests facilitates the testedbaseline being re-established in a very transparent way.

Test DataAssociated data collected during execution of the tests is usedas evidence to support the test result. This data should becross-referenced to the test numbers that were executed toproduce it.

Parent Ref Test Method Expected Result Actual Result Pass / Fail(Initials / Date)

F201 T1 Confirm that the system maintains a line velocity of 1 m/s in the The independentlytank outlet leg pipework by performing the following method: measured line velocity is1. Attach a calibrated flowmeter at the point... at least 1 m/s at the2. Ensure all user valves are open. tank outlet leg.3. Measure the flow rate over a period of...4. Attach a trace of the trended flow rate over the measurement

period.

F349 T2 Perform the following tests at the tank inlet line: The line velocity is at1. Attach a calibrated flowmeter at the point... least 1 m/s at the tank2. Ensure all user valves are closed. inlet line.3. Etc.

Table H. Test protocol format.



SummaryDocuments and their contents should be kept modular inaccordance with the V-model. Use of requirement referencenumbers aids requirement traceability up and down thevertical axis of the V-model as well as test traceability alongthe horizontal axis of the V-model.

By carefully structuring specification and test documentsand by using reference numbers, systems can be put in placeto promote the establishment and maintenance of the testedbaseline.

References1. GAMP® 4, Good Automated Manufacturing Practice

(GAMP®) Guide for Validation of Automated Systems, p.22, Figure 6.2: A Basic Framework for Specification andQualification, International Society for PharmaceuticalEngineering (ISPE), Fourth Edition, December 2001,www.ispe.org.

2. GAMP® 4, Good Automated Manufacturing Practice(GAMP®) Guide for Validation of Automated Systems,Appendix M3 - Guideline for Risk Assessment, Interna-tional Society for Pharmaceutical Engineering (ISPE),Fourth Edition, December 2001, www.ispe.org.

About the AuthorDavid Paspa, BE (Hons), is OperationsDirector at Synertec, a life sciences consult-ing company with offices in Melbourne,Sydney, and Singapore. He is a Director andTreasurer of the ISPE Australia Affiliate. Hehas worked in the pharmaceutical and bio-technology industries since 1990 on largeand small projects for the manufacture of

sterile and solid dose products, veterinary products, andAPIs. His involvement has been in the areas of softwarespecification, design, development, and commissioning withan emphasis on documentation and validation. Paspa alsohas been involved in process engineering activities, includingplant modelling, throughput optimization and de-bottlenecking. He has worked as both a client and a supplier.He can be reached at [email protected].

Synertec Pty. Ltd., 84 Johnston Street, Fitzroy 3065,Victoria, Australia.

Glove Integrity Testing


Safe Access using Glove Ports –Facts and Fictionby Johannes Rauschnabel, Albrecht Kühnle,and Kuno Lemke

This articledescribesprocedures totest gloves,shortcomings ofplastic glovebehavior, andglove fixation. Inaddition, itpresents a newimpulsetechnique for theimprovement ofreproducibilityand repeatabilityof a pressuredecay test.

Introduction

Barrier systems, such as aseptic isola-tors, Restricted Access Barrier Sys-tems (RABS) and glove boxes are beingused more and more in pharmaceuti-

cal production, research, and laboratories. Thepurpose of these barrier systems is to separatea process area from a surrounding environ-ment – either to shield the process from con-taminants coming from outside, or to shield theenvironment from hazardous products inside.If manual operator access is needed eitherduring processing, for maintenance reasons, orfor environmental monitoring inside the bar-rier, glove ports are needed.

The integrity of these glove ports is crucialfor maintaining sterility of the barrier system.Therefore, frequent inspection of the glove portsis required. The FDA guidance for industry onaseptic processing1 asks for “With every use,gloves should be visually evaluated for anymacroscopic physical defect. Physical integritytests should also be performed routinely,” whileEC GMP ² requires “Monitoring should be car-ried out routinely and should include frequent

leak testing of the isolator and glove/sleevesystem.” Thus, some kind of leak testing has totake place in addition to visual inspection.

ChallengeGlove leaks have many reasons: mechanicaldamage through contact with sharp equipment,tools or broken glass, which cause cuts in theglove material. Clamping gloves in mechanicalinstallations very often results in holes that areless discrete and hard to detect. The mostfrequent types of leaks come from heavy use ofgloves and aging. The commonly used glovematerial “Hypalon” has a layered structure,which tends to flake off over time, thereforecausing precarious perforations of the glovemembrane. The glove membrane do not onlyshow elastic behavior, it also has viscous prop-erties, especially if it is stressed by tension ormechanical load. These stressed areas have areduced membrane thickness and are moresensitive to mechanical impacts. Membranerupture is very often the consequence of overstressing gloves locally. Chemicals can have asimilar effect: some are capable of reducing the

tensile strength of the glove.Any type of leak should be

detected with a physical glovetesting procedure - at any posi-tion on the sleeve, cuff ring, orglove assembly.

The standard for operatingisolator systems is to work witha second disposable glove (80%of all users)³ – so that directcontamination is not very likelyeven in the case of a small leak.Additionally, in most isolatorsystems, an over pressure isapplied inside barrier isolators,which prevents ingress of air-borne contamination from out

Figure 1. Proportions ofa 10 μm leak in a 400μm membrane.





side to inside. But germs are able to grow through holes –against any pressure level. Bacterial spores have a diameterdown to 1 µm, while vegetative bacilli and yeasts grow todiameters of 10 µm and more. By above mentioned diam-eters, the acceptance criteria for a glove tester seems to be

given. Figure 1 demonstrates the proportions of a 10 µm holein a 400 µm thick glove membrane (~15 mil).

However, no existing physical test method is capable ofdetecting leaks in glove assemblies down to 1 µm diameter.What is worse: cuts of 100 micron scale may not be detectedin every position of the glove assembly.

Existing glove test devices for the pharmaceutical indus-try apply different procedures utilizing pressure differencebetween inside and outside of the glove, including (I) oxygenmeasurement in a nitrogen chamber, (II) air flow measure-ment, and (III) pressure decay measurement - Figure 2.

Other techniques such as the bubble test, ammonia test, orHelium leak test are useful to detect the leak location, but donot give quantitative results, which help to decide ‘passed’ ornot ‘passed.’ And conductivity measurement of a glove filledwith electrolyte in a tub with distilled water can’t be appliedfor in situ testing.

For test procedure (I), the glove is placed with a special cuffring into a vacuum chamber, which is evacuated and filled withnitrogen. If there is a leak, air (containing 20% oxygen) frominside the glove leaks into the chamber, where a gas sensormeasures the oxygen level. This level can be correlated to a leakrate. The procedure is sensitive due to high test pressure of4000 Pa and is very accurate (acceptance criteria is an oxygenconcentration of 500 ppm, which refers to an artificial 40 µmhole). Gloves can be tested during production, but the test is notable to challenge the complete glove assembly.

Test procedure (II) pressurizes the glove over a certaintime at about 600 Pa. The air volume per time needed tocompensate pressure loss by leakage is measured with a flowmeter and correlated to a leak rate. Theoretical calculationscome to a minimum acceptance criteria of 2 ml/min, whichwould correlate to a hole diameter of approximately 66 µm. Inpractical testing, this method can only achieve results downto minimum 100 µm diameter, but reproducibility is poor.The ‘history’ of the glove/sleeve assembly plays an importantrole (see below).

Test procedure (III) ‘pressure decay’ is the most commonphysical testing method with pharmaceutical isolators. Ac-cording to the ISPE 2004 isolator survey, ³ more than 70% ofresponses apply some kind of pressure decay testing. In thisprocedure, a positive pressure expands the glove (assembly).Either directly or passing a certain pressure level, the mea-surement starts. After a certain time, the end pressure istaken. From the pressure drop, a leak rate can be calculated.The principle of this test is very simple and can be performedwith a pressure gauge and a stopwatch only. But the down-side of this procedure could be lack of reproducibility.

All these procedures take as a basis the perfect ‘virtuallynew’ glove, which always behaves the same way duringmeasurement. But this is fiction and not reality.

During pressure difference testing, the following two pa-rameters can not be kept constant: air volume and gloveelasticity. Air is compressible, which has to be accepted andwill not vary with other parameters kept constant. A glove isnot a fully elastic system, but shows some plastic behavior.The glove expands non-proportionally to the pressure levelFigure 2. Procedures for glove testing applying pressure difference.



and does not keep its volume while keeping the pressureconstant. This behavior is dependant on glove material, glovethickness, age, and history of the glove. Results differ withnew gloves, heavily used gloves, freshly tested gloves, andaged gloves from stock, in addition to temperature fluctua-tions during measurement, which impact air compressibilityand glove elasticity. From all these shortcomings, physicalglove testing could be questioned principally.

Innovation and ResultsThere is one aspect, which gives a glimmer of hope forreproducible testing: the influence of this non-linear behav-ior decreases with stress level. This means that with higherpressure or by longer period of pressurization, the gloveconverges to a kind of ‘equilibrium state,’ from which glovebehavior could be taken as constant.

Higher pressure levels should help, but have their limits. A3000 Pa pressurization of a standard glove/sleeve assembly(400 µm membrane thickness) results in inflating the sleeve toballoon size while the glove keeps the shape. Therefore, thehigh pressure difference for test procedure I (4000 Pa) – whichis principally useful – can only be applied for the glove alone.

The alternative – prolonging the period of pressurization– is feasible for complete glove/sleeve assemblies, but couldtake hours, which is not very practical.

A new approach stresses the glove/sleeve assembly byrepeated pressure impulses, which are applied as soon as thepressure inside the glove drops below the starting pressurelevel (1200 Pa). The equilibrium state is achieved within 10 to20 minutes depending on glove/sleeve thickness and material.As shown in Figure 3, the impulse frequency reduces over time,which is an indicator for approaching the equilibrium state. Allfollowing results are achieved by that technique.

To demonstrate the influence of stressing time, glove/sleeve assemblies in two states (brand new, old/used) wereperformed - Figure 4. The results are the average value ofthree measurements. The longer the stressing time, thesmaller the deviation. The same can be observed for thepressure level: the higher the pressure, the better the repro-ducibility; 1200 Pa showed to be the optimum pressure levelfor standard glove/sleeve assemblies.

To demonstrate the repeatable equilibrium state at thebeginning of the measurement, a series of tests were per-formed with varied pause time between each run: from nopause to two hours - Figure 5. The results spread in a windowof 20 Pa, which is sufficiently accurate for reliable detectionof an artificial 100 µm hole with standard glove/sleeve com-binations.

A prototype of reinforced Hypalon sleeve combined with astandard Hypalon glove was tested. The results were verypositive: test pressure could be doubled (2400 Pa) withoutoverstretching the sleeve. Stressing time could be cut in half andresolution doubled allowing reliable detection of 50 µm holes.

Practical AspectsHoles of a 100 µm diameter or less can hardly be detectedduring visible inspections. Even cuts of a millimeter in sizecould remain undetected could remain undetected by visualcheck. Under unfavorable circumstances, the detection ofthese cuts also could be a challenge for physical glove testingprocedures. Depending on the location and orientation ofsuch cuts, the force induced by pressurization is either suffi-cient to open the leak or not. And it is not only the pressurelevel that affects opening probability, but also the direction ofthe cut in relation to the tension from pressurization - Figure6. Cut locations in the sleeve can easily be detected, but cutson the finger tip are much more difficult to be opened – thisis because of geometrical aspects (see above), and also due tothe higher glove thickness at the finger tip.

Another important aspect is the tightness of the completeassembly: port, sleeve, cuff ring, glove. The performance ofthe whole system is determined by the weakest link. Gloveand sleeve fixations are critical points. The very commonfixation by expanded o-rings clamping the glove (sleeve) on aring is not the way o-rings should be used. They have therebest sealing properties with being pressed between two facesof a connection. In case of oval shaped glove ports, the o-ringused as an expander shows a high contact pressure at thesmall radius sections compared with insufficient contactpressure at the big radius sections. What is more: Hypalonmaterial tends to crawl under mechanical load – such ascontact pressure by an o-ring – which has an effect on thetightness. Therefore re-adjustable fixations have advantages

Figure 4. Variation of stressing time with set of old vs. new gloves.

Figure 3. Measurement preparation utilizing impulse technique.



Figure 6. Behavior of leaks from needle prick in relation to leakorientation.

achieve reproducible results. Increasing the pressure level orprolonging the stressing time shows improvement in repro-ducibility of the test results. A new impulse technology helpsto save time. Recommendation from practical experienceemphasizes the need for a reliable glove fixation/sealingtechnique. Utilization of gloves in RABS, isolator, and con-tainment systems requires a slightly different approach, butit should be performed in conjunction with integrity testing.

Physical glove testing with a reliable procedure helps toreduce dependence from adherence to visual inspection SOPs.

References1. Guidance for Industry, Sterile Drug Products - Produced

by Aseptic Processing, US Department of Health andHuman Services, FDA, CDER, September 2004.

2. EC Guide to Good Manufacturing Practice, Revision toAnnex 1, Manufacture of Medicinal Products, EuropeanCommission, DG Enterprise, May 2003.

3. Lysfjord, J., and Porter, M., “Barrier Isolation History andTrends,” Pharmaceutical Engineering, Vol. 23, No.2, March/April 2003, pp. 58-64.

AcknowledgementsThe authors acknowledge support from students WolframSchindler and Matthias Bergmann performing many of thetests. We also thank Mr. J. Jackson for proof reading and Mrs.M. Biedermann for her support in graphics.

About the AuthorsDr. Johannes Rauschnabel is ProductManager at Bosch Packaging Technology.He is responsible for Bosch barrier systemsand the Bosch PharmaLab. He has morethan 14 years of experience in research anddevelopment. Dr. Rauschnabel graduated asa chemist from Eberhard-Karls University ofTübingen and holds a PhD in organic chem-

istry. He can be contacted by email: [email protected].

Figure 5. Repeated measurement with variated pause time betweenthe runs.

over expanding techniques involving o-rings.Experience on filler isolators show that many perforations or

leaks in glove/sleeve assemblies are caused by interventionswith stopping machinery. Routine work – such as environmen-tal monitoring in the isolator, transfers, and routine adjust-ments do not have impact on glove integrity as often. Leaks inthe sleeve are very common, typically coming from overstretch-ing (bad ergonomics), wear through leaning on the glove portring, and untrained or inappropriate handling by the user.

Hypalon is the favorite glove material for use in isolatorsbecause of its stability against oxidizing agents, such ashydrogen peroxide vapor. In case of Restricted Access BarrierSystems (RABS), which are installed inside cleanrooms ofhigh air quality (at least ISO 7), the gloves have to besterilized prior to transfer into the operations room. The costof glove/sleeve assemblies is very high – so for economicreasons, being able to sterilize glove/sleeve assemblies mul-tiple times can be an important requirement althoughautoclavability of Hypalon gloves is poor. The mechanicalproperties change after 6 – 8 autoclave cycles and result inleakage after 12 – 15 cycles. Alternative glove materials couldbe an option to overcome that disadvantage.

With containment systems, the risk of operator contamina-tion require a different glove approach. The need for mechanicalstability and leak tightness comes from GMP and HSE require-ments. For aseptic containments with positive pressure, leakscould blow contaminants into the operator area. Therefore,thicker gloves or two layered types can be the better choice. Inaddition to gloves with improved mechanical properties, singlepiece types can be recommended to reduce interfaces.

For containment operation in general, glove testing is animportant part of health and safety precautions. The mea-surement procedures described above are applicable, butthicker glove membranes require higher pressure levels todetect small leaks.

SummaryPhysical glove integrity testing is required by regulatoryguidance. Investigations demonstrate that gloves to be testedapplying pressure difference should be prepared in order to



Albrecht Kühnle is a Mechanical DesignEngineer at Bosch Packaging Technology.He has 10 years of experience in develop-ment of washing, sterilizing and filling pro-cesses, and machines for liquid pharmaceu-tical products. Presently, he is focused on thedevelopment of Bosch barrier systems,mainly on isolators and cRABS, as well as

appropriate components, such as transfer systems, gloveports, glove testers, etc. He graduated with a mechanicalengineering degree at FH Aalen. He can be contacted by e-mail: albrecht.kuehnle@ boschpackaging.com.

Robert Bosch GmbH, Blaufelder Strasse 45, D-74564Crailsheim, Germany.

Kuno Lemke is Design Engineer for BoschPackaging Technology. He has 33 years ofexperience in hygienic design of packagingmachinery with focus on aseptic applicationsfor food and pharmaceutical. He is involvedin development and advanced engineering ofequipment for sterilization and testing.Lemke graduated from the school of technol-

ogy at Ludwigsburg/Germany. He can be contacted by e-mail:[email protected].

Robert Bosch GmbH, Stuttgarter Strasse 130, D-71332Waiblingen, Germany.

Plant Access Control Solutions


The Case for cGMP Compliant PlantAccess Control Solutions forPharmaceutical Laboratories andManufacturing Areasby Walfried Laibacher

This articleconsiders thebenefits ofvalidating plantaccess controlsystems tosupportconsistentproduct qualityand to liftproductivityimprovement.

Introduction

Historically, validation of automationsystems has focused on informationtechnology and process control solu-tions. Building Management Systems

(BMS) were thought to be ‘no-impact systems’to product quality. How things have changed.

This is supported by new methods and toolsavailable for Good Manufacturing Practice(GMP) assessment. They, in turn, leverage USFood and Drug Administration (FDA) guide-lines for pharmaceutical companies adopting arisk-based approach to product quality. But,even before the “Pharmaceutical cGMPs for

the 21st Century” were announced,1 BMSs werecategorized as a one process control system2

type – one with direct impact upon drug qual-ity, and therefore, patient safety.

A multitude of regulated manufacturersaround the world are embracing Heating, Ven-tilation, and Air Conditioning (HVAC) environ-mental control solutions – one of several BMSapplications – in their inventory of GMP criti-cal computerized systems as a must for reduc-ing risk of non-compliance and business dis-ruption. GMP-relevant records typically con-firm temperature, humidity, and differentialpressure as well as particulate matter in criti-cal cleanroom, laboratory, and production envi-ronments.

Forward-thinking regulated manufacturersare increasingly turning their attention to plantaccess control systems as an integral part of theBMS – encouraged also by FDA’s 21st Centuryinitiative for risk-based management to usenew automated systems for enhancing the qual-ity of the product being manufactured.

Early adopters are looking to manage physi-cal access to their laboratories and manufac-turing facilities in much the same way. But howcan a plant access control solution impact drugquality and what drives these organizations?After all, it is not just a case of regulatorycompliance with GMP. The author, WalfriedLaibacher, a regular participant in ISPE’s Eu-ropean seminar program, argues that, inde-pendent of plant design, access control solu-tions may be increasingly considered as criticalto product quality, and hence, need complianceto regulatory requirements. But more than

Figure 1. Access CardTerminal with integrateddisplay (shown here inconjunction with theTime and Attendanceapplication, a securitymanagement functionwithin an integratedbuilding managementsolution.)





that, GMP compliant access controlsolutions constitute a key differentiatorin a complex, mission-critical environ-ment.

Access Control Solutionsas a Direct Impact System

When assessing the GMP impact ofautomated systems in pharmaceuticalplants, some typical questions come tomind:

• Does the system preserve the prod-uct status?

• Does it produce data to support prod-uct acceptance or rejection?

• Does the system control a process(e.g., a Distributed Control Systemor DCS) that can impact productquality?

• Is there independent verificationthat the control system is perform-ing as intended?

These same questions apply to auto-mated access control solutions.

It is likely that regulators are look-ing at the physical access to your plantareas. There are at least two pharma-ceutical companies in Southern Eu-rope that were inspected by their localregulatory body and asked about theirmethod of identifying and segregatingpeople accessing critical areas. Bothdecided to introduce computerized ac-cess control systems and rated them as“direct impact system.” Indeed, the FDAhas already issued a 483 on the subjectin the United States. Here is the ex-ample:

“... Controlled by an automated build-ing security system that functions withthe use of electronic key cards assignedto personnel... ...management reliesupon physical security to access thesystem. There are some concerns re-garding the software controls for thissecurity management system:

• ...functional design requirementsare not established…..

• ...design control documentation hasnot been established... (e.g., usergroups are not defined, specifiedconfigurations for individual usergroups are not defined)

• ...records of periodic review of thepersonnel assigned to each usergroup are missing

• ...there is no Standard OperatingProcedure (SOP) to assign any oneuser to a specific user’s group on thesystem.

Not only is the FDA criticising the lackof system specification, it is taking anoperational perspective and raisingawareness of the potential risk of non-qualified people being able to accessthe system. This, inevitably, courts risk.Upon entry operators could add users,grant them access to non-controlledcritical areas, and deactivate vital se-curity configurations - anti pass-backmechanisms for example.

A risk analysis of current businesspractices would have shown this phar-maceutical manufacturer’s access con-trol system to be vulnerable. This, inturn, may have prompted a change toits classification - from non-GMP rel-evant to GMP-relevant. Without aspecification detailing how the systemshould operate, this manufacturercould not provide sufficient qualifica-tion evidence. Failure to do so put regu-latory compliance and product qualityat risk.

Why is Access Controlso Vital?

The following discussion will reviewthe GMP-impact assessment processof computerized systems in pharma-ceutical plants.

Any regulated manufacturer has avariety of automated solutions, all ofthem formally listed in an inventorytogether with their respective valida-tion approaches. This provides theframework for the validation planningprocess and determines the classifica-tion of such systems into GxP-criticalor GxP-non-critical.

Applying the GMP-Impact Assess-ment questionnaire to access controlsystems invariably prompts an affir-mative response to one or more issues.This confirms its categorization as adirect impact system – one that ac-cords qualification for confirmation ofregulatory compliance.

Here are some ways in which accesscontrols impact security and why theyare so critical to preserving productintegrity:

• Poor management of access controlrisks unauthorized personnel gain-ing entry to storage rooms and tam-pering with the product or vital in-gredients. An access control systemwould govern access permissions tocritical zones within the pharma-ceutical plant. Only those with cur-rent permission to enter would beable to do so.

• Physical permissions to critical ar-eas require current SOP trainingcertification. ‘Hygienic require-ments in cleanroom zones’ call forappropriate protective suiting forexample. The two inspected Euro-pean pharmaceutical companiesmentioned before also were askedhow they can assure that peopleaccessing critical areas are trainedabout procedures in place for those.Integration of access controls withan electronic training managementsystem will deny employee access tothis area in the event of lapsed orunavailable training records. In-stead of granting entry to a quali-fied drug production area, the keycard reader display would stop ac-cess, flag up the reason why, andrecommend the necessary remedialaction. In our example it might say:“Entry prohibited. No training onSOP Hygienic Instruction, Rev 1.2.Please contact site training admin-istrator.” See Figure 1 for AccessCard Terminal showing individualmessages.

• Robust audit trails, supported bydocumented evidence, provide fur-ther proof of entry and egress. Theygive additional assurance that the



process consistently operates in ac-cordance with its predefined speci-fication – in other words, validationensures the automated systemworks as originally intended.

These are just three examples, butthere are many other opportunities forGMP compliant access control systems.Some pharmaceutical manufacturersvideo record door entry points to criti-cal production areas. Resulting foot-age can support a security system, pro-viding for a visual check of SOP deploy-ment – real time and/or historically.Such electronic video records are usedby a pharmaceutical manufacturer inthe UK to demonstrate compliance withSOP deployment on hygienic instruc-tions – ready for GMP inspections. Se-curity management solutions can evendeliver alert notifications of anythingimproper such as the wrong color cloth-ing or attempted entry at an unsched-

uled time.Intelligent access control systems

also can be used to accept or reject aproduct.

Not only will they alarm in the eventof unauthorized entry to a critical plantarea, the transit log report will providedocumented evidence supporting prod-uct acceptance/rejection. As an ex-ample, it turned out that an alreadydismissed employee has a criminalbackground. Transit log reports helpedidentify to which areas this person hadaccess. This is particularly importantgiven the increasing incidence of phar-maceutical counterfeiting.

Furthermore, smart access controlscan govern entry to areas housing di-rect impact systems such as DCS orEnterprise Resource Planning (ERP) -another good reason for rating them asa direct impact system.

But the argument doesn’t stop there.Qualification supports the use of

integrated building and process man-agement solutions by making use ofthe same electronic key cards for ac-cessing rooms and DCS login.

Alerts are generated automaticallyin the event of the DCS being accessedby a key cardholder fraudulently usinga second key card to make systemchanges under a different name. Eventhough this second card would, in prin-ciple, give access to the process controlsystem, the DCS login is denied be-cause the corresponding key owner hasnot physically entered the control room.Verification between integrated sys-tems only serves to enhance security ina regulated manufacturing environ-ment.

Returning to the principles of GMPimpact assessment, a single ‘yes’ in theGMP assessment questionnaire doesnot necessarily require you to qualifythe whole access control system. Sys-tem boundaries focus qualification and

Figure 2. Typical system architecture for a plant access control system (software configuration at all three levels).



validation effort on the mission-criti-cal components of the direct impactsystem. This would not, for example,include a car recognition subsystemgranting access to the parking lot.

ISPE’s Baseline® Guide on Com-missioning and Qualification3 providesa useful guide and process for deter-mining system boundaries at the plan-ning phase.

How Best to Achieve acGMP Compliant Access

Control System?ISPE’s Good Automated Manufactur-ing Practice Guide (GAMP®) for thevalidation of automated systems – isthe ‘de-facto’ industry standard for vali-dation of automated systems. It is rec-ommended by the FDA as an effectivetool for these purposes.

A proven GAMP-based validationapproach for validating HVAC environ-mental control systems (as one part ofthe BMS) can be directly applied toaccess control solutions. This is becauseaccess control architecture follows thesame three-layer model as for environ-mental management - Figure 2.

• Supervisory Level: the level ofpresentation (status and alarms),human-system interface (applica-tion parameters, system configura-tion, and system control), and his-torical data management (events,values, and transits).

• Peripheral Level: the level of de-cision-making and feedback, it com-prises controller devices specializ-ing in the management of specificapplications.

• Field Level: the level at which thesystem interacts with the externalworld (employees, visitors, gates,and detectors for example). It ismade up of readers, displays, key-boards, actuators, and digital sen-sors.

On the environmental side, there aretemperature, humidity, and pressuresensors controlling, for example, an airhandling unit. With access controls,there are key card readers or any bio-metric devices and door contacts.

For both systems, the applicationlogic resides at the peripheral level.The management application on thesupervisory level may even be the samein state-of-the-art, integrated buildingmanagement solutions.

As discussed in the ISPE Baseline®

Guide on Commissioning and Qualifi-cation, components of all three levelsare obviously identified as critical com-ponents requiring qualification: thecard or biometric readers (palm reader,retinal scans, or others), followed bythe application configuration in theindependent working controller of theperipheral level. Finally, the setup toconfiguration, analysis, and reportingmeans provided by the supervisory levelalso require qualification, having docu-mented evidences about the systemsetup that only authorized operatorscan perform changes to the access con-trol system and that such changes arerecorded to support audit trail require-ments. The system should have provenmechanisms in place to reject elec-tronic record tampering and may havebeen configured so that critical controlactions can just be performed by elec-tronic signature means as defined in21 CFR Part 11. This just gives anexample which emphasizes that quali-fication has to address the whole sys-tem far beyond the visual system com-ponents like key card readers or bio-metric devices.

GAMP groups software into five dif-ferent categories, from category one,describing the validation approach ofsoftware type ‘Operating Systems,’ tocategory five, ‘custom or bespoke code.’

HVAC environmental control sys-tems, DCS or SCADA systems fall intocategory four software, ‘configurablesoftware package’ (though specific busi-ness operations may vary from thisassessment). A computerized plantaccess control system also can be as-signed to this group.

From this, it follows that a lot ofcommon practices for HVAC hardwareand software qualification can be du-plicated and applied directly to accesscontrols. As example, transit manage-ment, key card management for usersand zones, Present-in-Zone, and anti-passback configuration are functions

that come to mind in respect of Opera-tional Qualification (OQ).

Keeping the Validated Statein the Operational Phase

Qualifying access control solutions forcompliancy is relatively straight-for-ward.

In the context of an environmentalcontrol system, for example, set pointchanges on room temperature will avoidcostly ‘out-of-specification’ wastage ina production environment; they willforce adoption of control parameters inline with SOP on change control. Cor-rect change control deployment is vitalto staying GMP compliant.

However, access control systemsrarely require change control deploy-ment. For the most part, this only comesinto play in the event of system expan-sion. Rather, day-to-day actions in theoperational phase - visitor manage-ment for example – are recorded in theevent buffer for audit trail.

There are obviously less changes tosuch a plant access control systemneeded which makes it much easier tomaintain the validated operationalstate.

Access Control SolutionsCompliant to

21 CFR Part 11?The ability to qualify an access controlsolution enables pharmaceutical manu-facturers to leverage further produc-tivity improvement in their day-to-daybusiness operations. In line with FDA21 CFR part 11 guidelines it allows forElectronic Records and Electronic Sig-natures (ER&S), and with this, elec-tronic data reports for regulatory in-spection.

It also satisfies other crucial FDA21 CFR part 11 criteria including thefollowing:

• generation of exact, timely, andtamper-proof records

• Chronological audit trails, anotherkey aspect of compliance. Synchro-nization with a master clock on anorganization’s IT network is onesolution guaranteeing that allevents and transits are in the same



time reference as all other companysystems.

• Training. Although usually coveredunder operational procedures, theaccess control solution can help toensure that only trained personnelmonitor and configure the system.This necessitates operators beingassigned a profile that contains in-formation on their security level andarea assignments. Classification inthis way can be used to manageoperators; to restrict them to seeingand controlling only those parts ofthe system for which they are cur-rently trained. Their scope can bechanged, on-line, as new trainingoccurs. Operator permission can beconfigured down to the smallestdetail. Within a given area for ex-ample, an operator might be able tomonitor and configure certain pro-cedures. Timely refresher coursesalso can be managed in this way,thus, preserving product integrityand ensuring optimal plant uptime.

These are just few items to be consid-ered in the scope of ER&S usage inaddition to the application examplesusing ER mentioned in previous sec-tions.

The recently released GAMP® GoodPractice Guide4 on ER&S is now avail-able to help pharmaceutical firmsaround the world to identify their criti-cal e-records. In seeking to provide afocused approach on this subject, itfollows the more efficient top-downprinciple on critical evaluation of e-records by applying a risk based ap-proach rather than cost intensive, bot-tom-up evaluation of computerized sys-tems. With that, it follows the FDA’s21st Century initiative for risk-basedmanagement and considers and com-plies with international regulations.Being aligned to the principles pre-sented in the 2004 FDA Part 11 Guid-ance, its recommendations are wellthought-out.

Today non-GMP;Tomorrow GMP

Fact: things change. New legislationand the interpretation of it according

to industry guidelines and currentthinking is constantly evolving. Newguidelines point to regulatory direc-tion and provide support for improvingcurrent operational practices. Eventhough a pharmaceutical manufacturermay currently assess its access controlsystem as a non-GMP system, bestpractice may change as the ‘c’ in cGMPcomes into play.

What does this mean in the contextof a plant access control solution?

If it is assessed as a “no-impactsystem,” then the supporting specifica-tions, installation and commissioningspecifications need not stand up to thestringent criteria for qualification docu-mentation. Should, in time, it becomeGMP classified as a result of integra-tion with another direct impact com-puter system for example, a retrospec-tive validation might be appropriate.This would provide the necessary dataand information needed to support sys-tem documentation and the requisitevalidation as very clearly outlined inthe PIC/S Guidance5 for inspectors ofcomputerized systems in regulated“GxP” environments.

But retrospective qualification is nosmall task. More often than not, it ismore expensive than implementing aprospective validation framework. Ex-perience shows that many companiesfaced with such a dilemma chose to up-grade - even replace - their existing com-puterized system rather than backtrack.

Stay Ahead of the CurveIt is common knowledge that the regu-latory bodies regard plant access con-trols as a direct impact system in cer-tain manufacturing environments.This requires validated evidence of com-pliance.

Mandatory or not, early adopters ofbest practice are staying ahead of thegame and driving performance gain byvalidating their automated access con-trol systems. Indeed, putting a GAMP-conforming validation process in placenot only ensures that your access con-trols are working as intended, it deliv-ers big business rewards. Not only doesit prove regulatory compliance, in thelong run, it cuts costs dramatically andprotects product integrity.

In short, it makes good businesssense. A sound prospective validationapproach to project delivery and opera-tion is infinitely preferable to cumber-some and often complex remediationby retrospective validation. A proac-tive stance also gives regulated manu-facturers the opportunity to implementelectronic records and signature andlift productivity improvement still fur-ther.

There are plant access control solu-tions available today that can do this;that are ready to meet the require-ments for 21/CFR part 11 compliance.Not only do they give pharmaceuticalmanufacturers a perfect understand-ing – and record – of their access con-trol processes, they bring about re-duced risk of business disruption. Theyenable pharmaceutical companies to:

• stay ahead of the curve by attendingto their peripheral systems today

• be proactive – to be an early adopterand leader of best practice

• guarantee product quality and regu-latory compliance – and, from this,strengthen competitive advantage

It’s in your hands.

References1. FDA, “Pharmaceutical cGMPs for

the 21st Century: A Risk-BasedApproach,” http://www.fda.gov/cder/gmp/index.thm

2. GAMP® 4, Good Automated Manu-facturing Practice (GAMP®) Guidefor Validation of Automated Sys-tems, Chapter 9.3 - Types of ProcessControl Systems, International So-ciety for Pharmaceutical Engineer-ing (ISPE), Fourth Edition, Decem-ber 2001, www.ispe.org.

3. ISPE Baseline® Pharmaceutical En-gineering Guide, Volume 5 - Com-missioning and Qualification, In-ternational Society for Pharmaceu-tical Engineering (ISPE), First Edi-tion, March 2001, www.ispe.org.

4. GAMP® Good Practice Guide: A Risk-



Based Approach to Compliant Elec-tronic Records and Signatures, In-ternational Society for Pharmaceu-tical Engineering (ISPE), First Edi-tion, April 2005, www.ispe.org.

5. Pharmaceutical Inspection Coopera-tion Scheme (PIC/S), Good PracticesFor Computerized Systems In Regu-lated “GxP” Environments PIC/SGuidance, PI 011-1, 20 August 2003,http://www.picscheme.org.

About the AuthorWalfried Laibacherholds a diploma in ap-plied sciences for elec-trical engineering. Heis currently the Vali-dation Service Lead forHoneywell BuildingSolutions in Europe,

Middle East and Africa. Based in Ger-many, he oversees Honeywell’s spe-cialist validation sales and engineer-ing teams, ensuring that they applycommon validation practices for envi-ronmental conditions to the company’sgrowing pharmaceutical customerbase. This includes on-site project sup-port, recently as the validation con-sultant in a multi-million Euro processcontrol project for a major Europeanbiotech company. Since joiningHoneywell in 1988, Laibacher has heldseveral positions in software develop-ment from software engineer to projectleader for HVAC and building man-agement solutions in an SEI CMMlevel 3 certified organization. As a mem-ber of the German speaking GAMP D-A-CH Forum, he contributes to a GAMPSIG on “Cooperation Models betweenUsers and Suppliers.” He is a regularon the European lecture circuit, hisspeciality being new-generation vali-dation solutions and integrated build-ing automation systems and how theysatisfy pharmaceutical customer needsfor compliant environmental control.He can be contacted by e-mail:[email protected].

Honeywell GmbH, HoneywellBuilding Solutions, SeligenstädterGrund 11, D-63150 Heusenstamm,Germany.

Options Analysis


The Staging Option and DrugDevelopmentby Neal Lewis, PhD, David Enke, PhD, andDavid Spurlock, PhD

This articleinvestigates thestaging optionto analyze anexisting casestudy thatinvolves thepotentiallicensing of adrug compoundthat is indevelopment. Itdemonstrateshow optionsanalysis is auseful tool inadding insight tothe decisionmaking processwhenconventionalvaluationmethods are notdecisive.

Introduction

R esearch and Development (R&D)projects are routinely evaluated to de-termine if the projects are feasible andworthy of continued funding. Most

R&D organizations have more ideas than theyhave resources to fund them so projects mustcompete for available resources, includingmoney and talent. A widely used technique forevaluating projects is Discounted Cash Flow(DCF). In this method, the Net Present Value(NPV) is determined by discounting forecastedfuture cash flows by a required rate of return,as shown in equation 1.

T FVTNPV = -I0 + Σ __________ (Eq1)

(1 + r)T

whereI0 is the original investmentFVT are the future cash flowsr is the interest rateT is the time increment

The discounted cash flow method is widelyused to determine the value of projects, and hasbeen widely embraced by industry. Despite itswide use, discounted cash flow biases evalua-tors toward conservative conclusions. Goodideas are sometimes not pursued because themethod provides an NPV that is often too low.1

Management usually has flexibility during thecourse of R&D projects, and this flexibility is

not accounted for in the DCF technique.2

Projects with NPVs that are very high areconsidered good investments from the DCFperspective. Projects with NPVs that are nega-tive are generally abandoned because they willnot deliver the required return. Projects withNPVs close to zero require significant addi-tional effort to determine if such projects shouldbe funded or abandoned. Real options analysiscan be used to add insight to the funding deci-sion, especially when DCF analysis finds anNPV that is close to zero. Real options analysisoffers an alternative that determines a valuefor managerial flexibility and provides an Ex-panded Net Present Value (ENPV).

OptionsA financial option is an asset that gives theowner the right, without an obligation, to buyor sell another asset (such as a quantity ofcorporate stock) for a specified price at orbefore some specified time in the future. A realoption is a potential investment, such as aproject, that is funded only if the firm decidesit is in its best interest to do so. The option toinvest in a project (or not to invest) has value.In real options analysis, the option to invest inthe project creates an ENPV, which is definedas:3,4

ENPV = NPV + Option Value (Eq2)

When NPV is quite large, the option value willnot have a significant impact on the decision:the NPV signals that the project is worthy ofinvestment. When NPV is very negative, eventhe best option values will not be large enoughto create a positive ENPV, and the projectshould not be pursued. If the future cash flowsare known with certainty, then the DCF tech-nique should be used. Real options have theirbest use under conditions of uncertainty, andwhere management has the ability and the

Variable Financial Options Real Options(such as stock options) (such as projects)

T Time to expiration Time to expiration

r Risk-free interest rate Risk-free interest rate

X Exercise price Implementation cost

S Stock price PV of future cashflows

σ Volatility of stock Volatility of futureprice movement returns

Table A. Optionvariables.



Options Analysis


Figure 1. Davanrik decision tree.

willingness to exercise its flexibility. The option value placesa price on the value of this flexibility, and the ENPV identifieshow much the firm should be willing to pay to keep the project(or option) open.

Real options analysis is based on the mathematics offinancial options, and has received widespread attention andacclaim since the early 1990s. Few companies have extensiveexperience with real options. However one notable authorfeels that real options will replace NPV as the central methodfor investment decisions in the future.1

There are five primary variables involved in the optionvalue calculation for financial assets. The Black-Scholespricing model estimates the value of a simple call option (C)based on the current stock price (S0), strike price (X), volatil-ity (σ), risk-free interest rate (r), and the time to expiration(T). The equation is:

C = S0N(d1) - Xe-rTN(d2) (Eq3)

where

S0 σ2

(1n ______) + (r + ______) TX 2

d1 = __________________________σ √T

__d2 = d1 - σ √T

N(dx) is the cumulative standard normal distribution of thevariable dx.

The five variables of financial options have direct equiva-lents in real assets - Table A.3,5 Note: of the five variables usedin real options analysis, four are used to calculate NPV andare usually available to the analyst. The new variable thatneeds to be considered is the volatility of the project’s futurerate of return.

The value of simple options can be quickly calculatedusing the Black-Scholes model, but the math becomes verycomplex if the option becomes more complicated. Binomiallattices also can be used to determine the value of financialand real options, and is a preferred method for complexoptions. This technique is explained later.

Compound options are those options that are dependenton the value of other options.6 Compound options can besequential, and are sometimes called staging options. Manyprojects are funded in phases; good project managementencourages this approach. If there is a phased investment,and succeeding investments are dependent on previous

investments, then a sequential compound (staging) optionmay exist.7 Virtually all drug development projects arephased investments, and most are sequential compoundoptions.8

The Costs of Clinical TrialsIn 2003, the average cost for testing and successfully launch-ing a new drug was estimated at $900 million.9 While thereare significant costs involved in launching a new product,much of the total cost is associated with clinical trials. Thecosts of a clinical trial include trial design, patient recruit-ment, clinician cost, product cost, monitoring, data analysis,close out and reporting results, coordination with regulatoryauthorities, and administrative costs.10 Antibacterial trialsare particularly expensive to run, and can cost $50,000 perpatient; one Phase III trial can cost half a billion dollars.11

Clinical trial costs vary depending on the type of drug, thenumber of people involved in the trial, and the product claimsthat are trying to be proven. Increasingly, marketing claimsare driving additional clinical trials in order to ‘position’ aproduct in the marketplace.12

Case StudyThe following hypothetical case study was published as anexample of evaluating a drug development opportunity.13 Thecase has been used in several business schools as an exampleof how to use decision trees to perform valuation studies.

A small pharmaceutical firm has developed a new chemi-cal compound that they believe has a good chance of becominga prescription drug. The chemical is called Davanrik, and hasthe potential to treat both depression and obesity. The smallfirm lacks the resources to complete the approval process,and so has approached a large pharmaceutical company witha proposal to license the chemical and complete the drugdevelopment process. At the time of the licensing offer,Davanrik was finishing pre-clinical development and was

Table B. Probability of success for drug approval.

Phase Average Chance of Success, %

Average(14) Small Molecule(10) Large Molecule(15)

Preclinical 35 --- ---

Clinical Phase I 75 73 75

Clinical Phase II 50 45 50

Clinical Phase III 70 --- 73

Approval 90 --- 81

Options Analysis


getting ready to enter clinical testing.The testing and approval process was expected to take

seven years. If all went according to plan, Davanrik wouldhave 10 years of exclusive marketing rights, beginning withthe New Drug Application (NDA) approval. The typical prob-abilities of success for each clinical phase are known in thedrug industry; these are shown in - Table B. One of the largestrisks in the drug industry is the clinical development process.The value of a drug development project is determined in partby evaluating the probabilities of success at each stage of theclinical process.

During clinical testing, Davanrik would be given to 20 – 80healthy people to determine human safety.16 The testing wasexpected to cost $30 million and take two years to completewith an estimated 60% chance of success.

During clinical testing, the chemical would be given to 100– 300 people to determine the efficacy for treating depressionand/or weight loss. The probability of success for the depres-sion indication was estimated at 10%, the probability ofsuccess for the weight loss indication was estimated at 15%,and the probability of both indications being successful wasestimated at a 5% probability. Phase II testing was expectedto require two years to complete, and would cost $40 million.

In Phase III clinical testing, Davanrik would be given to1000 – 5000 people to determine safety and efficacy in a broadspectrum of the population. This testing was expected to takethree years to complete and depended on successful resultsfrom Phase II. If the earlier testing demonstrated that thechemical was effective only for depression, then the Phase IIItrials would cost $200 million and have an 85% chance ofsuccess. If Davanrik were found to be effective for weight lossonly, then the trials would cost $150 million and have a 75%chance for success. If Davanrik were found to be effective forboth, then the cost of Phase III would be $500 million with a70% chance of success.

Davanrik has the potential of generating large profits. Ifthe drug were approved only for depression, it would cost$250 million to launch, with a present value of future cashflows of $1.2 billion. If Davanrik were approved only forweight loss, it would cost $100 million to launch with apresent value of future cash flows of $345 million. If thechemical were approved for both depression and weight loss,it would cost $400 million to launch with a present value offuture cash flows of $2.25 billion. All costs have already beendiscounted to the present time. While the development costsare high and the chances of success are low, the potentialpayout is very high if success can be achieved. The questiontherefore becomes: Should Davanrik be licensed?

The Decision Tree and Traditional ValuationFigure 1 shows a decision tree for the Davanrik problem. Thetree starts in year zero with the beginning of Phase I; thisphase lasts two years and costs $30 million. The probabilityof success is 60%. If Phase I is successful, then Phase II maybegin. Phase II lasts two years and will cost $40 million. Thechance of success for Davanrik as a depression medication is10%; the chance of success for weight loss is 15%; and the

chance of success for both is 5%. This relatively low probabil-ity of success is in line with industry norms.10 If any of theseare successful, then Davanrik may enter Phase III testing.Each of the indications has its own cost and probability ofsuccess. If Phase III is successful, then the product may belaunched, pending FDA approval.

We can analyze this information using traditional meth-ods. The most commonly used valuation method is NPV.Using the potential income, potential costs, and the probabil-ity of each, a Net Present Value can be determined for eachindication. First look at costs, beginning at time T = 7 years.Normally, costs would need to be discounted back to timezero, but the costs have already been discounted. It is as-sumed that payments are made at the conclusion of eachphase. Assuming Phase III is successful, we have two costs atT=7: the Phase III cost and the product launch cost. ThePhase III clinical study will need to be paid for, but theproduct will be launched only if it is successful. For thedepression indication, this includes $250 million for thelaunch and $200 million for the Phase III study. At yearseven, the probability-adjusted costs are (Figure 1):

Depression cost = 200 + (.85)(250) = $412.5 in Year zerodollars

Weight loss cost = 150 + (.75)(100) = $225Both cost = 500 + (.70)(400) = $780

Assume that the $40 million cost of the Phase II clinical studyis divided equally between the three possible indications. Atyear four, the probability-adjusted costs are:Depression cost = 40/3 + (412.5)(0.1) = $54.58Weight loss cost = 40/3 + (225)(0.15) = $47.08Both cost = 40/3 + (780)(0.05) = $52.33

At year two, the end of Phase I, the probability-adjusted costsare:

Table C. NPV results.

Indication Discounted Discounted Net PresentIncome Cost Value

Depression 58.14 42.75 $15.39 million

Weight Loss 22.13 38.25 - $16.12 million

Both 44.89 41.40 $3.49 million

Figure 2. The Binomial lattice.

Options Analysis


Table D. Davanrik variables.

Depression Weight Loss Both

PV of the future cash flows 1200 345 2250

Royalty paid (5%) 60 17.25 112.5

Effective PV of future cash 1140 327.75 2137.5flows, S

Launch cost, X4 250 100 400

Phase 3 cost, X3 200 150 500

Phase 2 cost, X2 13.33 13.33 13.33

Phase 1 cost, X1 10.0 10.0 10.0

Time for the launch (years) 7 7 7

Time for Phase 3 (years) 7 7 7

Time for Phase 2 (years) 4 4 4

Time for Phase 1 2 2 2

r (risk-free interest rate) 5% 5% 5%

Estimated volatility 40% 40% 40%

Table E. NPV and options analysis results.

Indication NPV ENPV

Depression only 15.4 740.18

Weight loss only -16.1 120.3

Both depression and weight loss 3.5 1367.8

Depression cost = 30/3 + (54.58)(0.6) = $42.75 millionWeight loss cost = 30/3 + (47.08)(0.6) = $38.25 millionBoth cost = 30/3 + (52.33)(0.6) = $41.40 million

Because all of these costs were already discounted to yearzero, these are also the probability-weighted costs for thethree options at year zero.

The income also can be determined using decision trees.The income streams have already been discounted so we do notneed to adjust for time, only probability. The income streamwill occur only if all tests are successful so the probability isbased on success of all clinical trials. A royalty of 5% isassumed, decreasing all future income to a factor of 0.95.

Depression income = (1200)(0.6)(0.10)(0.85)(0.95)= $58.14 million

Weight loss income = (345)(0.6)(0.15)(0.75)(0.95)= $22.13 million

Both = (2250)(0.6)(0.05)(0.70)(0.95)= $44.89 million

We can determine the NPV by subtracting the costs from theincome - Table C.

Strictly speaking, the NPV technique indicates that theproject should be undertaken for depression and for the dualindication, but not weight loss. If the weight loss indicationwere not pursued, the costs that were assumed by the weightloss indication would need to be shifted to the other options,increasing their costs. At year four, the probability adjustedcosts then become:

Depression cost = 40/2 + (412.5)(0.1) = $61.25Both cost = 40/2 + (780)(0.05) = $59.00

At year two, the end of Phase I, the probability-adjusted costsare:

Depression cost = 30/2 + (61.25)(0.6) = $51.75Both cost = 30/2 + (59.00)(0.6) = $50.40

The Net Present Value then becomes:

Depression NPV = 58.14 – 51.75 = $6.39Both cost = 44.89 – 50.40 = -$5.51

The NPV for “Both” is not viable, making it so that thedepression indication must assume all Phase I and Phase IIcosts. When the NPV for the depression indication alone iscalculated in a similar way, the result is -$20.61. The recom-mendation would be that the Davanrik licensing agreementshould not be pursued.

Even the most generous of organizations would have con-cerns over funding this project. The payback method, still usedby many organizations, shows that the project will not pay itscosts until late in the project’s life. The Internal Rate of Return(IRR) analysis shows that the project might return the weightedaverage cost of capital under the best of conditions, but notmuch more. Under any traditional test method, if the weightloss indication is not pursued, the other projects cannot affordto cover the fixed costs of the early clinical testing. This is ascenario where NPV is close to zero, and management spendssignificant time gathering information in order to make thebest possible decision. An advocate of this project will need toproduce a different rationale to justify moving forward. Op-tions analysis could help such an advocate.

The Davanrik Projectas a Sequential Compound Option

The value of the real option can be calculated with thebinomial options approach, using a lattice to demonstratealternative possibilities over time.17 The starting point is thepresent value of the future cash flows (S0). Over time T, twoconditions can result at each decision point: one positive upoutcome and one negative down outcome (hence the termbinomial). Over several time steps, we can create a lattice asshown in - Figure 2.

The option valuations require a risk-free rate of return.For the binomial lattices to work correctly, costs will becompounded at the risk-free rate so that standard time valueof money equations can be used for discounting. In effect, thecosts and incomes will be compounded at 5% so that they canlater be discounted at the same 5%.

In order to calculate the option value using binomiallattices, the variables need to be identified. For the Depres-sion indication:

Present value of future cash flows, PV = 1200Royalty paid (5%) = (1200)(0.05) = 60Effective PV of the future cash flows, S = 1200 – 60 = 1140Cost of product launch, X4 = 250Cost of Phase III, X3 = 200

Options Analysis


Cost of Phase II, X2 = 40/3 = 13.33Cost of Phase I, X1 = 30/3 = 10.0Time for the launch and for Phase III = 7 yearsTime for Phase II = 4 yearsTime for Phase I = 2 yearsRisk-free interest rate r is assumed to be 5%Volatility must be estimated

There are several ways to estimate the volatility of projects,but these methods will not be investigated in this article.18

Merck generally begins an analysis based on a volatility of40%; 19 and we will use the same.

We need to determine a few variables that are used to solvethe binomial lattice. The up step (u) is defined as:

___

u = eσ√δT (Eq4)

where δT is the change (δ) in time (T) for the step.The binomial lattice is constructed so that each time-step

is equal to one year, so δT = 1.__

u = e0.4√1 = 1.492 (Eq5)

The down step (d) is defined as 1/u,

d = 1/u = 0.670 (Eq6)

Instead of discounting our cash flows at a risk-adjusted rate,as is often done in discounted cash flow analysis, we candetermine a risk-neutral probability and then discount ourcash flows at a risk-free rate. This is the standard techniqueused for building binomial lattices as they are used in valuingoptions.20 The risk-neutral probability p is

erδT – d e0.05*1 – 0.670p = ________ = ________________ = 0.4637 (Eq7)

u – d 1.492 – 0.670

The variables for the other indications can be calculated ifthey are not given. A summary of the input variables is shownin - Table D.

The option value for the depression indication may becalculated using the technique for sequential compoundoptions.20 In this case, we have four sequential options. PhaseI clinical testing occurs first, and future work may notcontinue without success in Phase I. Therefore, Phase II isdependent on the successful completion of Phase I. Similarly,Phase III is dependent on Phase II, and product launch is

dependent on Phase III (as well as FDA approval). Due to thecomplex nature of this sequence, the binomial lattice methodis the simplest method for the calculation of the option value.The calculation consists of five lattices, each related to theprevious one.20 Microsoft Excel is used to speed the calcula-tions. The first lattice is the underlying lattice, starting withthe value of S on the left ($1140 million). This is shown inFigure 3 with time steps of one year. Each step is calculatedwith an up-step being

___ ___

u = eσ√δT = e.4√1 = 1.492

The up-step is 1.492 times the previous value, and a down-step is 1/u, or 0.670 times the previous value.

The next lattice is the equity lattice for the execution of theproject - Figure 4. For this lattice, the cost incurred for thelaunch of the project is subtracted from the value at yearseven (shown in the right hand column - Figure 3). This formsthe basis for the new right-hand column - Figure 4. Thesuccessive columns to the left are then discounted by theequation

Vt-1 = [(p)(V+) + (1-p)(V-)]e-rδT (Eq7)

whereVt-1 is the value in the next column to the leftp is the risk-neutral probabilityV+ is the upper value (the up-step)V- is the lower value (the down-step)r is the risk-free interest rateδT is the length of the time step

Figure 3. Depression underlying lattice. Figure 4. Depression equity lattice at Launch.

Figure 5. Depression equity lattice at Phase 1.

Options Analysis


The entire lattice is completed from right to left. Given thevalues in year seven, the top number in the lattice for year sixis Vt-1 = [(0.4637)(18392.13) + (1-0.4637)(8068.76)]e-(0.05)(1) =12228.96.

The next lattice is the equity lattice for Phase III of theproject. This lattice is not shown by itself, but can be seen inthe top right corner of the five-lattice diagram of - Figure 6.For this lattice, the cost incurred for the third clinical phaseis subtracted from the value at year seven in the previous

lattice, creating the new right-hand column. The successivecolumns to the left are discounted using the same risk-neutral probability.

The next lattice is the equity lattice for Phase II of theproject. Again, this lattice is not shown by itself but can beseen in Figure 6. The cost incurred for the second clinicalphase, which occurs in year four, is subtracted from the valueat year four in the previous lattice. This lattice will have thesame values as the previous lattice for years five, six, and

Figure 6. Five-lattice binomial calculation, depression indication.

Options Analysis


seven. A new year-four column is created and the columns tothe left are again discounted as before.

The final lattice is the equity lattice for Phase I of theproject - Figure 5. For this lattice, the cost incurred for one-third of the first clinical phase, which occurs in year two, issubtracted from the value at year two in the previous lattice.This lattice will have the same values as the previous latticefor years three through seven. A new year-two column iscreated, and the columns to the left are calculated as before.The far left hand column of this last lattice represents thevalue of the option in year zero, and provides the value of theoption. This is also the ENPV of the project. The ENPV for theDepression indication is $740.18 million.

The value for this option is clearly higher than the NPVcalculation of $15.39 million. This is due to several factors.First, the option never achieves a value below zero. If condi-tions indicate that the value would be negative, then the option(or project) would not be executed and the value is simply zero.This represents the value of management’s flexibility to notfund a money-losing project. Second, this is a high-risk projectwith a very large potential payout. Whereas NPV decreasesthe value of the project when risk is present, options analysisincreases the value when risk is present. The chances of havinga positive decision are enhanced with options analysis.

The primary question at this point is whether it is worth-while to fund the beginning of this project. For the Depressionindication, the project needs to justify the expenditure of $10million for its fair share of Phase I clinical testing. Managersare not being asked to fund the entire project, only the firstphase. Based on the option analysis, this method says thatproject expenditures of up to $740.18 million are justified.

Similarly, the option values for the other indications canbe calculated. A summary of the results is shown in - Table E.For all indications, the ENPV is substantially higher than thepreviously calculated NPV.

The ENPV gives a clear signal that the project should beundertaken. Real Options Analysis was able to improve thedecision where NPV analysis was not clear. The projectshould be continued if the Phase I clinical trials are success-ful. If they are not successful, then the project should beabandoned. The project should again be evaluated beforePhase II money is committed. The complete five-lattice bino-mial calculation is shown in - Figure 6.

Implications for thePharmaceutical Engineer

The valuation of a project is an aspect of project managementthat can be crucial to the success of a project. Valuation isdiscussed extensively in the academic literature and in thepopular business press. The issue is relevant to anyone whois attempting to justify a project. Valuation is also relevant tobusiness accounting and finance, and is an important part oftax law. Discounted cash flow analysis is widely used in thepharmaceutical industry, and engineers need to be aware ofthe problems that these methods present. Discounted cashflow undervalues many projects, whereas the use of realoptions helps to determine a more accurate project value.

A key advantage in the use of the staging option is that itassumes that future costs will be spent only if it is in the bestinterest of the firm to do so. A future phase will be undertakenonly if the previous phase is successful. This is a muchdifferent assumption than NPV, where future costs are ad-justed only by their estimated probability of occurrence.

Options analysis has been criticized for being a “blackbox.” Many managers do not understand the methods and donot understand how a given option value is calculated. Thebinomial lattice approach is a flexible technique that is basedon simple mathematics. The method is easy to learn and canbe understood by a wide range of interested parties. Ad-vanced mathematics such as calculus is not needed whenusing the binomial lattice approach. This method helpsalleviate the “black box” mentality that has hindered theapplication of options analysis in the past.

Real option values are determined based on a set of fore-casts, including future cash flows and future costs. It istherefore necessary to realize that option values are estimates,and are only as accurate as the input variables. When neces-sary, sensitivity analysis can be applied to option analysis.

The staging option is one of several related real optionstools available to the pharmaceutical engineer. Projects aresometimes delayed until additional information can be ob-tained. In this case, a deferral option may be a useful analysistool. Projects are abandoned for a variety of reasons, in whichcase an abandonment option could prove useful. Projects canbe expanded or contracted depending on market success, andthe expansion option or the contraction option can be used tobetter define the worth of such projects. While these optionsare beyond the scope of this article, they are analysis toolsthat can aid in project valuation.

ConclusionsStaging options can be used to provide a more accurateproject valuation when NPV is not decisive. By approachingprojects with a staged investment strategy, we limit theinvestment and the risk at the early stages. This also pro-vides time for better understanding of the future cash flowsand costs, allowing managers to make better decisions asuncertainty is resolved and project outcomes are more clearlydefined. This approach can be used in a variety of industries,and is very applicable to medium and large projects. It isespecially useful in the pharmaceutical industry where drugdevelopment is required to be a staged investment.

Drug development can often be viewed as a sequentialcompound option. As an option, funding of a developmentproject first hinges on the decision to fund the first round oftesting. The entire project does not need to be funded at onetime. Management has the option of either funding or aban-doning the project at a later date, and this option has value.Early stages require a relatively small initial investmentcompared to the potentially large future funding require-ments so options analysis may alter the decision of whetherto fund the project. Most of the information needed to performthe options analysis is the same as would be used to deter-mine a project’s NPV with the exception of volatility. The

Options Analysis


volatility of the project’s future returns must often be esti-mated, and this volatility can be difficult to determine.Furthermore, calculation of the expanded NPV is more com-plex than determining the traditional NPV, but the binomiallattice approach provides a value without the use of complexmathematics. The calculation method using the binomiallattice has been demonstrated, and shown to be a relativelystraight-forward method for calculating option value.

The staging option is an important approach for all kindsof projects since much of the project work is performed usingstaged funding. The demonstrated case study clearly showshow the final decision can change when viewed from anoptions perspective.

References1. Copeland, T. and V. Antikarov, Real Options; A Practitioner’s

Guide, New York: Texere, 2001.2. Miller, L.T. and C.S. Park, “Decision Making Under Uncer-

tainty – Real Options to the Rescue?” The EngineeringEconomist, Vol. 47, No. 2, 2002, pp. 105-150.

3. Trigeorgis, L., Real Options: Managerial Flexibility andStrategy in Resource Allocation, Cambridge, Massachu-setts: The MIT Press, 1996.

4. Contractor, F.J., Valuation of Intangible Assets in GlobalOperations, Westport, Connecticut: Quorum Books, 2001.

5. Schweihs, R., The Economic Analysis of Real Option Value,1999. Available: http://www.Willamette.com/pubs/insights/99/realoption.html.

6. Geske, R., “The Valuation of Compound Options,” Journalof Financial Economics, Vol. 7, No. 1, 1979, pp. 63-81.

7. Benaroch, M., “Option-Based Management of TechnologyInvestment Risk,” IEEE Transactions on Engineering Man-agement, Vol. 48, No. 4, 2001, pp. 428-444.

8. Herath, H.S.B., and C.S. Park, “Multi-Stage Capital In-vestment Opportunities as Compound Real Options,” TheEngineering Economist, Vol. 47, No. 1, 2002, pp. 1-27.

9. Goldfisher, A., “Rising Drug Costs,” Venture Capital Jour-nal, Vol. 44, No. 10, 2004, p. 55.

10. Webster, J., and T. Philippon, “Valuation of BiotechnologyCompanies and their Assets,” Life Sciences Quarterly,Deloite & Touche, LLP, 2nd and 3rd Quarter, 2004, pp. 20-28.

11. Sellers, L.J., “Big Pharma Bails on Anti-Infectives Re-search,” Pharmaceutical Executive, Vol. 23, No. 12, 2003, p.22.

12. Langreth, R., “Drug Marketing Drives Many Critical Tri-als,” Wall Street Journal, Vol. 232, No. 97, 11/16/98, p. A10.

13. Ruback, R.S. “Merck & Company: Evaluating a Drug Li-censing Opportunity,” Harvard Business School Case Study,December 19, 2001.

14. Kennedy T., Pharmaceutical Project Management, NewYork: Marcel Dekker, Inc., 1998.

15. Witzke, David, “Biotechnology: Biotech Platforms – LatentValue,” Morgan Stanley Research, October 2002.

16. Kellogg D. and J.M. Charnes, “Real-Options Valuation fora Biotechnology Company,” Financial Analysts Journal,Vol. 56, No. 3, 2000, pp. 76-84.

17. Dixit A.K., and R.S. Pindyck, Investment Under Uncer-tainty, Princeton, N.J.: Princeton University Press, 1994.

18. Lewis, N. and D. Spurlock, “Volatility Estimation of Fore-casted Project Returns for Real Options Analysis”, Proceed-ings of the National Conference of the American Society forEngineering Management, 2004.

19. Nichols, N.A., “Scientific Management at Merck: An Inter-view with CFO Judy Lewent,” Harvard Business Review,January-February 1994.

20. Mun, J., Real Options Analysis, Hoboken, N.J.: John Wiley& Sons, 2002.

About the AuthorsNeal Lewis is an Associate Professor in theCollege of Information Technology and Engi-neering at Marshall University. He is thecoordinator of the MS program in technologymanagement in the division of Applied Sci-ence and Technology. He earned his BS inchemical engineering and PhD in engineeringmanagement from the University of Missouri

- Rolla, and an MBA from the University of New Haven. He hasmore than 25 years of industrial experience with Procter &Gamble and with Bayer Corporation. He can be contacted byemail: [email protected] or by telephone: 304-746-2078.

Marshall University, College of Information Technologyand Engineering, 100 Angus E. Peyton Dr., South Charleston,West Virginia 25303.

David Enke is an Assistant Professor in theEngineering Management Department at theUniversity of Missouri-Rolla (UMR). He is theDirector of the Laboratory for Investment andFinancial Engineering and is a member ofUMR’s Smart Engineering Systems Lab, In-telligent Systems Center, and Energy Researchand Development Center. He received his BS

in electrical engineering and MS and PhD degrees in engineer-ing management, all from UMR. His research interests are inthe areas of financial engineering, financial risk, financialforecasting, investment, engineering economics, and intelli-gent systems.

David Spurlock is an Assistant Professor inthe Engineering Management Department atthe University of Missouri-Rolla. He earned aPhD in organizational psychology from theUniversity of Illinois at Urbana-Champaignand has more than 10 years of engineering andmanagement experience in the industry. Hisresearch interests include individual and group

judgment and decision making processes; managing people inorganizations; organizational change, organizational develop-ment and program evaluation; and the influence of technologi-cal change on workplace behavior.

University of Missouri-Rolla, Dept. of Engineering Manage-ment, 1870 Miner Cr., Rolla, Missouri 65409.

Reprinted from

PHARMACEUTICAL ENGINEERING®

The Official Journal of ISPE

November/December 2005, Vol. 25 No. 6

Country Profile - India


This feature inPharmaceuticalEngineering is

designed so thatyou can tear itout, three hole

drill (if desired),and keep it with

other CountryProfiles as they

are published.

Look for theCountry Profileon Thailand in

the January/February issue of

PharmaceuticalEngineering.

Dear ISPE Members,

On behalf of the ISPE India Affiliate, I am pleased to present the profileof the Indian pharmaceutical industry as it has evolved over the pastfew decades.

A profile of the healthcare system in India would be incomplete withouta mention of the two renowned ancient Indian ayurvedacharyas,Sushruta and Charaka. Indians have always been health and medicineconscious giving the Indian pharmaceutical industry a strong base.Roots are very strong.

The profile also discusses how the Indian industry has developed in thelast 50 years. The proactive functioning of various associations andbodies, which have enhanced the growth of the Indian pharmaceuticalindustry, also deserve mention.

The profile also looks at the regulatory systems in India at present witha focus on the Drugs Controller General of India (DCGI) and his role inprice control of various preparations. The aim in the Indian setting isto provide effective drugs and affordable prices. Hence today thepharmaceutical industry in India is faced by numerous challenges. Butthere are many opportunities as well. The Indian pharmaceutical sectoris fast being recognized by international markets and is expected toemerge as the global driver in the coming years. Research anddevelopment, a key factor for progress of any drug industry, also hasbeen discussed in sufficient detail.

With strong roots, and a proactive present, India has tremendous futurepotential. India has taken a significant step forward in achieving thegoal of becoming a globally competitive market in the 21st century.India is also taking fledgling steps into the global arena by forgingcontacts with research and marketing based companies. Alliances withboth companies and institutions like ISPE are giving Indian companiesan opening into the world’s global research networks and the chanceto gain access to new technologies like Process Analytical Technology(PAT).

We hope this profile succeeds in its intention to give a contemporaryoverview of the pharmaceutical industry in India.

Yours truly,

Ajit SinghPresidentISPE India Affiliate



A Look at the Pharmaceutical Industryin Indiaby Shabbir Badami, Pharmacist, ACG Worldwide and RajshriSrinivasan, Free-lance Journalist and Copywriter

The History

The healthcare industry in India is one of theoldest known to mankind and is steeped in tradi-tion. Ayurveda (Sanskrit: ayu: life; veda: knowl-

edge of) or ayurvedic medicine is a comprehensivesystem of medicine based on a holistic approach that ismore than 6,000 years old. This art of healing had beenheld in high esteem in ancient India. It was elevated toa divine status and Dhanvantari the practitioner ofthis art was defined as the God of Medicine. Evenordinary practitioners of this art the Ashwinikumars- were given a special status in mythology and folklore.

Two early texts (from centuries BC) of Ayurveda arethe Charaka Samhita and the Sushruta Samhita.They outline remedies for practically every conceiv-able condition starting from cancer, diabetes, choles-terolemia to fever, pain, and depressive disorders. Themain objective of Ayurveda is removing the cause ofillness with little or no side effects and not just curingthe symptoms. With such strong roots, Indians havealways been health and medicine conscious, thus,putting the pharmaceutical industry on a firm footingfrom ancient times.

A Country of ContrastsIndia is a country of contrasting cultural and economicbackgrounds. Because of the vast cultural and eco-nomic diversities, disease profiles and subsequentlytreatment measures are very distinct and disparate.Diversities in India are not restricted only to culture,language, and tradition, but are also reflected in lifestylepatterns. Hence these affect the way the pharmaceuti-cal market behaves. The pharmaceutical industry inIndia manufactures drugs and medicines to cater tothe needs of a variety of disease conditions for patientsfrom various economic strata of society. Medicines are

accessible at registered retail outlets, in ur-ban as well as rural areas, covering everycorner of the country.

This huge contrast poses challenges as well asopportunities to the pharmaceutical sector. They arelike two faces of one coin, two poles of one magnet. Thechallenges and opportunities that the Indian pharma-

ceutical industry is encountering currently have neverbeen so gigantic. Nevertheless, because of its large andyouthful human resource, low-cost, diversity, and de-mocracy, it is becoming evident that this century is‘India’s Century.’ It is not only that most of the jobs arecoming to India; even for most of the developed coun-tries it is essential to keep attracting Indian humanresources for their betterment.

The Pharmaceutical IndustryToday, the pharmaceutical industry in India is in thefront rank of India’s science-based industries withwide ranging capabilities in the complex field of drugmanufacture and technology. It holds a leadershipposition in the third world in terms of technology,

Continued on page 4.



quality, andrange of medi-cines manufac-tured. Fromsimple headachepills to sophisti-cated antibioticsand complex bio-

technology derived products, almost every type ofmedicine is available in India. The industry has about10,000 active units, out of which 300 are in large andmedium sectors. More than 400 Active Pharmaceuti-cal Ingredients (APIs) and more than 60,000 formula-tions in 60 therapeutic categories are manufacturedhere.

The organized sector of the pharmaceutical industryhas played a key role in promoting and sustainingdevelopment in this vital field. International and In-dian companies associated with this sector have stimu-lated, assisted, and spearheaded this dynamic devel-opment in the past 57 years and helped to put India onthe pharmaceutical map of the world. Several Multi-National Companies (MNCs) produce a large volumeof the world’s production in India. The value of thepharmaceutical market in India was US $5 billion in2003. It grew by 5.1% over 2002. Globally, the rankingof the Indian pharmaceutical industry is 4th in volumeterms and 14th in value terms. Its share of the globalpharmaceutical market is 1.8% in value and 8% involume terms. The differential is largely due to themuch lower market selling prices in India.

The pharmaceutical units in India are hi-tech hi-GMPand the industry has the largest number of US-FDAapproved units outside the US, while the number ofDrug Master Files filed with the US FDA was 126,higher than Spain, Italy, China, and Israel put to-gether in 2003. In addition, the Indian pharmaceuticalsector is mature and is supported by a strong manufac-turing base.

For the year 2004, the Indian pharmaceuticalindustry registered an annual turnover of US$5.97 billion with a growth rate of 6.4%. Exportsaccounted for US $4.06 billion with a growth of10.2%. 300 bulk drugs were manufactured and thebulk drug production was valued at US $2.08 billion.More than 60,000 formulations were manufactured in60 therapeutic categories. The OTC market was val-ued at US $0.93 billion and the alternative medicinemarket touched US $0.97 billion. The export of bulkdrugs and formulations in 2003 were to the extent ofUS $3.1 billion. The US $0.051 billion vaccine marketis growing at the rate of 20% annually.

2004 saw six additional companies added to the US$0.55-1.11 billion bracket - Table B.

The Indian pharmaceutical market has some uniqueadvantages. The country has a solid legal frameworkand strong financial markets. More than 9000 compa-nies are publicly listed. It has a good network of world-class educational institutions and established strengthsin IT. The country is now committed to an open economyand globalization. Above all, it has about 200 millionmiddle class individuals with growing entrepreneur-ial spirit. India has the third largest English speakingscientific and technical manpower in the world. Forthe first time in many years, the international phar-maceutical industry is finding great opportunities inIndia. The process of consolidation, which has becomea general phenomenon in the world pharmaceuticalindustry, has started taking place in India. With itsrich scientific talents and research capabilities, Indiais well set to mark its place as the sunrise industry.

The Indian pharmaceutical industry is a success storyproviding employment for millions. It has a pool ofpersonnel with high managerial and technical compe-tence, as well as a skilled workforce. The employmentis provided directly by the organized and small scareunits or indirectly through the trade and ancillaryindustry. The organized sector, which comprises ofMNCs and indigenous companies, employs 0.29 mil-lion individuals, while the small scale units hire 0.17million persons. The area of distribution and tradeoffers job opportunities to nearly 1.65 million personswhile the ancillary industry hires 0.75 million per-sons. Hence the pharmaceutical industry offers jobopenings for nearly 2.8 million persons in the country.

The Pharmaceutical andAllied Industries

The ancillary industry is also very well developed. All

A Look at the Pharmaceutical Industry in IndiaContinued from page 3.

Type Quantity

Medicine 302

Chemistry 1,799

Biological Science 221

Analytical Technique 90

Table A. India’s patent count.

Turnover in US$ Billion Number of Companies

2002 2003 2004

>1.11 6 6 5

0.55 - 1.11 16 15 21

0.208-0.55 19 21 20

0.113-0.208 33 32 33

0.577-0.113 31 36 32

Table B.



manufacturing equipment and machineries are lo-cally available. While some excipients continue to beimported, India is self sufficient and well developed inthe ancillary industry requirements.

A full range of pharmaceutical manufacturing equip-ment is locally produced as are glass bottles and vials,empty two-piece capsules, as well as soft capsules,blister and packaging films, aluminum and laminatedtubes, and almost all other requirements for a fastgrowing vibrant formulation and bulk drug industry.India is important among the world leaders in thevolume of production of most such items. India is animportant exporter to both advanced as well as devel-oping countries.

Research and DevelopmentInvesting in R&D, Contract Manufacturing,and Biotech ResearchIndia’s track record and development, particularly inthe area of improved cost-beneficial chemical synthe-sis for various drug molecules, is excellent. The indus-try offers tremendous opportunity in the area of re-search and development without compromising qual-ity. The R&D in itself is emerging as a rapidly growingindustry with an estimated annual revenue of US$1.25 billion. There are almost a dozen major compa-nies engaged in innovative research to produce NewChemical Entities (NCEs). The R&D expenditure isUS $0.26 billion, 4% of sales (some companies arespending 6% of their sales on R&D).

Similar is the case of contract manufacturing. Becauseof economics, a genetically diverse population, a sig-nificant number of qualified providers with expertisein conducting and supervisingclinical trials in accordance withglobal standards, India is emerg-ing as an important hub for Con-tract Research and Manufactur-ing with patenting and certifica-tion activities assuming a promi-nent position in the country. In thechanging scenario, acquisitions

and mergers are no longerrare.

The biotechnology industryin India is in the nascent stages.About 85% of the fast growing bio-technology market is contributedby six products, all of which are

manufactured in India for captive consumption. Amongthese are Erythropoeitin, Hepatitis B vaccine, GrowthHormones, Interferons, and Insulin. The total biotechmarket in India is estimated to be US $1.4 billion.About 60% of this market is accounted for bybiopharmaceuticals; however, today, this market isgaining increasing importance.

Regulation of MedicinesAbout five decades back, when the Indian pharmaceu-tical industry was in its early stages, it dependedheavily on imports. Today, the industry has evolvedtremendously. It is at a level where it is almost com-pletely self-sufficient and has also developed a sizableexports market globally. The regulatory apparatushas evolved along with the industry. While industryoccasionally reports red-tape and over-regulation, in ahuge diverse and democratic country like India, bothregulation and distribution present their own chal-lenges.

The industry has a well-evolved regulatory system anda well-defined legislation, which operates on the prin-ciples of licensing all activities related to drug manu-facturing distribution and sales. It has defined sys-tems to monitor the import of drugs, new productintroductions, and good clinical and laboratory prac-tices. The legislation also provides separate rules forpsychotropic and narcotic drugs. The Indian Pharma-copoeia (IP) is an independent body, the work of whichis looked after by the pharmacopoeial commission.Ayurvedic, siddha, unani, and homoe-pathic medi-cines have separately defined legislation and GMPrequirements. The legislative requirements are con-tinuously reviewed and upgraded due to fast and

A Look at the Pharmaceutical Industry in India

Company Market Share (%) Growth (%)

2002 2003 2004 2002 2003 2004

Cipla 5.25 5.51 5.51 15 11 7

GlaxoSmithKline 5.92 5.64 5.44 -2 1 3

Ranbaxy 4.63 4.70 4.48 7 7 2

Nicholas Piramal 3.39 3.41 4.25 9 6 3

Sunpharma 2.92 3.08 3.29 18 11 14

Dr. Reddy’s 2.80 2.64 2.43 20 -1 -2

Zydus Cadilla 2.41 2.45 2.42 17 7 5

Aristo 2.10 2.19 2.32 12 10 13

Abbott 2.32 2.29 2.31 7 5 7

Alkem 2.26 2.17 2.18 12 1 8

Table. Top ten companies in the Indian market.Continued on page 6.



complex growth and with an eye on evolving interna-tional requirements.

The law defines products under price control and drugrelated advertisements are monitored through theDrugs and Magic Remedies Act.

The pharmaceutical industry has quality producersand many units are approved by regulatory authori-ties in the US, UK, and other advanced countries.India has the largest number of US FDA approvedmanufacturing facilities outside of the US. About 70units have been cleared by US FDA and nearly 253units have an approval from well-recognized interna-tional agencies. A large number of our new modernizedunits meet global standards. Newer and innovativedosage forms have been developed.

The prime mover in influencing the operating environ-ment of the pharmaceutical industry is thegovernment’s drug policy. Currently, the Drug Policyof 1986 as modified and rationalized in 1994 is in force.The Drug Price Control Order (DPCO) outlines theclassification of controlled and decontrolled productsand methods of price fixation and revision. The DPCOhas a three-tier control: on bulk drugs, formulations,and overall profitability. At present, there are 74 drugsunder price control (40% of the retail market).

The National Pharmaceutical Pricing Authority(NPPA) monitors the fixing and revising of the pricesof a number of drugs. Policy makers have realized theneed for giving incentives for R&D to the Indianpharmaceutical industry. They also believe that the

A Look at the Pharmaceutical Industry in IndiaContinued from page 5.

import liberalization process will assist theindustry to further improve its quality and com-petitiveness.

Future PotentialThe Indian pharmaceutical industry is mounting upthe value chain. From being a pure reverse engineer-ing industry focused on the domestic market, theindustry is moving toward basic research driven, ex-port oriented global presence, providing a wide rangeof value added quality products and services. Govern-ment policies will play an important role in definingthe future of the pharmaceutical industry. The effect ofproduct patent regime coming into effect from January2005 is being watched with interest.

The Indian pharmaceutical companies have been do-ing extremely well in developed markets such as theUS and Europe, notable among these being Ranbaxy,Dr. Reddy’s Labs, Wockhardt, Cipla, Nicholas Piramal,and Lupin. Such companies have their strategies inplace to leverage opportunities existing in formula-tions, bulk drugs, generics, novel drug delivery sys-tems, new chemical entities, and biotechnology. By theyear 2010, the R&D revenue is expected to touch US$4.89 (The Boston Consulting Group) while projectionfor the pharmaceutical market is US $25 billion (TheMcKinsey report).

With a prodigious knowledge pool and skilled man-power expertise in process chemistry and proven lead-ership in the field of IT, India has the necessaryingredients to become a dominant player in pharma-ceuticals.



Moreover, the Indian pharmaceutical industry is pass-ing through a wave of consolidation with the objectiveof strengthening brand equity and distribution inwhat is essentially a branded-generics market. In theperiod 1995-98, the Indian pharmaceutical industrywitnessed as many as 20 mergers, acquisitions, andtakeovers.

In the coming years, India is poised to become a majorplayer in the global healthcare industry. Qualitycoupled with the availability of technical and non-technical manpower availability makes India the idealglobal center for pharmaceutical outsourcing.

Sources• OPPI data• IMS report• IPMMA Newsletter• Ernst and Young report• ISPE Annual Meeting• Interview with Mr. Ashwini Kumar, Drugs Control-

ler General India• Interview with Dr. Venkateswarlu, Deputy Drug

Controller General India• Interview with Mrs. Bhavna Shah• ICRA Industry watch series - The Indian Pharma

Industry• ISPE India Affiliate Web site - www.ispeindia.org• IDMA, OPPI, and BDMA Web sites

About the AuthorsShabbir Badami is a pharmacist and works with thecorporate marketing team of ACG Worldwide. He hasbeen associated with the Indian pharmaceutical in-dustry for more than two decades and has worked withseveral multinational and Indian companies likeSchering Plough (Fulford), Zydus Cadila, and IntasPharmaceuticals. He has been responsible for brandpromotion and marketing activities within India, aswell as on a global level, giving him broad experiencein the pharmaceutical industry.

Rajshri Srinivasan is a free lance journalist andcopywriter. In her career span, she has interviewed

eminent medical and pharmaceutical person-alities and edited and written articles for

financial, medical, and pharmaceutical jour-nals and newspapers. During her time in the

pharmaceutical industry, she also has trained fieldstaff on various aspects of disease and drug profiles.Her work has taken her to nearly all parts of India.

A Look at the Pharmaceutical Industry in India

Visit the ISPE IndiaAffiliate Web site at

www.ispe.org/india/ forlocal events and updates.

ISPE India AffiliateBoard Members (Trustees)

Mr. Ajit SinghACG Worldwide

Mr. Gopal NairGrasp Enterprise

Dr. P.G. ShrotriyaM. J. Biopharma Pvt. Ltd.

Dr. A.K. SingalPharmaplan (India) Ltd.

Dr. M. VenkateswarluCentral Drugs Standard Control

Organisation

Mr. R.P. LalaKlenzaids

Mr. J. L. SipahimalaniCMA Laboratories

CONTACT US!

ISPE India Affiliatec/o ACG Worldwide

10th Floor, Dalamal HouseNariman Point

Mumbai 400 021 IndiaTel: 91-22-2287-2557

E-mail: [email protected]

ISPE Asia Pacific Office73 Bukit Timah Road#03-01 Rex HouseSingapore 229832Tel: 65-6330-6755Fax: 65-6336-2263

E-mail: [email protected]



The Indian pharmaceutical industry is represented by various associations prime amongthem are: the Organization of Pharmaceutical Producers of India (OPPI), the Indian

Drug Manufacturers’ Association (IDMA), and Bulk Drug Manufacturers’ Association(BDMA). These bodies effectively represent the pharmaceutical industry and issues facingit today. Another association that represents the industry are the trade associations RetailDrug Manufacturers’ Association (RDCA) which represents and tackles issues facing thetrade, retail, and distribution outlets. The Indian Pharmacy Association (IPA) represents allthe aspects of pharmacy and pharmacy education including hospital pharmacy, communitypharmacy, institutional pharmacy, and retail pharmacy.

• • •

Pharmaceutical Associations andOrganizations in India

Indian DrugManufacturersAssociation (IDMA)

Head Office (Mumbai):102-B, Poonam ChambersDr.A.B.Road, WorliMumbai 400 018 IndiaTel: 91-22-24944624 or 91-22-24974308Fax: 91-22-24950723Email: [email protected] [email protected]

Delhi Office:2nd Flr, B-4/115Safdarjung EnclaveNew Delhi 110 029 IndiaTel: 91-11-6171367Fax: 91-11-6171369E-mail: [email protected]

http://www.idma-assn.org

IDMA was founded in 1961. To-day, it is a premier association ofthe Indian pharmaceutical indus-try and has come to be regarded ingovernment, media and industrycircles as the Voice of the NationalSector. IDMA members compriselarge, medium, and small compa-nies from all over India, manufac-turing bulk drugs and formula-tions. IDMA plays a vital role inthe growth and development of

the industry, by taking up with thegovernment major issues such asprice control, patents, and trademark laws, quality and GMP, R&D,Exports, etc. and promoting betterunderstanding with the consumerorganizations, the press and othermedia on problems faced by theindustry.

Bulk DrugManufacturersAssociation (BDMA)

C-25, Industrial EstateNear SBHSanathnagar, Hyderabad500 038 A.P. IndiaTel: 91-40-23703910 or 91-40-23706718Fax: 91-40-23704804E-mail: [email protected]

http://www.bdm-assn.org

The BDMA India was formed in1991 with Hyderabad as its head-quarters. This is an all India-bodyrepresenting all the bulk drugmanufacturers of India. The Asso-ciation works for the consolidationof gains of the industry and servesas a catalyst between the govern-ment and the industry on the vari-ous issues for the growth of theindustry.

Organization ofPharmaceutical Producersof India (OPPI)

Peninsula ChambersGround FloorGanpatrao Kadam MargLower ParelMumbai 400 013 IndiaTel: 91-22-24918123,91-22-24912486, or91-22-56627007Fax: 91-22-24915168E-mail: [email protected]

http://www.indiaoppi.com

The OPPI was first established on27 December 1965 with Dr. HomiR. Nanji as the first President.This association was set up withthe aim of promoting developmentof the pharmaceutical industry inIndia. The first annual report waspublished on 31 December 1966.

Today, OPPI is a premier vibrantand knowledgeable organization.Its scope of activities include OTC,animal health products, biotech-nology among others. OPPI mem-bers (70 in number) manufacture300 bulk drugs. OPPI membersadhere to the code of pharmaceuti-cal practices

Testing GxP Systems


Key Principles and Considerations forthe Testing of GxP Systemsby the ISPE GAMP® Testing SIG

This articlepresents keyprinciples ofGxP criticaltesting includingtest objectivesand businessbenefits.

The finalpublishedGAMP® GoodPractice Guidemay differ fromthis article as aresult of theISPE externalreview.

Introduction

The GAMP® Testing Special InterestGroup (SIG) has developed guidelineson the testing of computer and soft-ware based systems that impact prod-

uct quality, patient safety, or patient confiden-tiality in the regulated healthcare industry.The GAMP® Good Practice Guide: Testing ofGxP Systems aims to provide users and suppli-ers with guidance on the following commonlyasked questions by those responsible for thetesting of GxP systems:

• What should I test?• How much testing is enough?• How should I conduct tests?• How should I document my testing?

Specifically, this GAMP® Good Practice Guide(GPG) is intended to take the principle of risk-based validation (as established in GAMP® 4)and provide practical advice on the applicationof this principle in the planning and executionof risk-based testing.

The intended audience for the GAMP® GPG:Testing of GxP Systems includes:

• users (responsible for the testing of GxPapplications)

• suppliers (responsible for the testing of stan-dard software and systems used in the regu-lated healthcare industries and for the test-ing of customized software developed forspecific users)

• systems integrators (responsible for the con-figuration of the product into a specific ap-plication which may contain custom code)

The GAMP® GPG has been written by users andsuppliers primarily associated with the phar-maceutical industry. This is reflected in someof the terminology defined and used through-out. However, the principles and guidance givenmay be of equal relevance in other sectors of theregulated healthcare industry, such as medicaldevices, biotechnology, biomedical, andhealthcare.

Finally, in the preparation of the GAMP®

GPG, the members of the SIG have deliberatelychosen not to redefine general principles oftesting best practice used in the wider softwaredevelopment and testing community. Focus isgiven to the practical application of these prin-

Figure 1. The use of riskassessment in determingthe scope and rigor oftesting.



Testing GxP Systems


ciples to the testing of GxP systemsand references for further reading areprovided, where appropriate.

This article presents an extract fromthe GAMP® GPG, including test objec-tives and business benefits, and thekey principles of GxP testing:

• use of risk assessment• testing in the life cycle• testing strategies• testing and hardware/software cat-

egories• testing responsibilities – supplier

and user

This article includes an overview ofuser and supplier responsibilities andconcludes with an overview of the con-tent of the GAMP® GPG. The GAMP®

GPG: Testing of GxP Systems is ex-pected to be placed on the members’area of the ISPE Web site at the end of2005 in electronic format.

The Objectives of TestingFor users, the basic underlying reasonfor testing a computer-based system orapplication is to provide a high level ofassurance that the system is fit for itsintended use, prior to the system beingused in the live environment.

For suppliers, the basic underlyingreason for testing is to prevent thepresence of avoidable defects in thesupplied system.

There are a number of different rea-sons why this is desirable, which maybe summarized as:

• assuring the safety of users, con-sumers (patients), and members ofthe general public

• increasing user confidence in thesystem

• reducing the cost of on-going sup-port

Business BenefitThe principle business benefit fromtesting systems is that it is more costeffective to move into the live environ-ment with systems that are fit for pur-pose.

Anyone who has been involved in aproject with insufficient or inappropri-ate testing learns that those problemsexposed only after testing are usuallythe most time consuming and trouble-some to resolve.

Where there is pressure to imple-ment systems in timescales that areunrealistic, there are several potentialknock-on effects:

• reduces the effectiveness and effi-ciency of the system at ‘go live’

• increases the maintenance and sup-port costs

• requires a costly program of correc-tive actions to be implemented, tocorrect faults, and meet the originalrequirements

• at worst, roll out a system that doesnot meet the basic user require-ments

The net effect is to increase the overallcost of implementing and owning the

system and to delay or prevent theeffective and efficient use of the sys-tem.

RegulatoryTesting is a fundamental requirementof current good practice with regard toachieving and maintaining regulatorycompliance. Although the need to testcomputer systems is defined by certainregulations, the way in which com-puter systems should be tested is notdefined in detail in specific regula-tions. However, supporting guidancedocuments issued by various regula-tory agencies suggest good practice in anumber of critical cases.

The nature and extent of computersystems testing should be defined andjustified on a system-by-system basis,and this may be based upon a docu-mented risk assessment. However, itis a basic regulatory expectation thatGxP computer systems require somedegree of testing.

Failure to test may undermine anyvalidation case and the compliance sta-tus of the system. Where discoveredduring regulatory inspection, this maylead to citations and warning lettersbeing issued and possibly a failure togrant new drug/device licenses, licensesuspension, or products being placedon import restrictions.

These regulatory expectations arebased on the basic principle that com-puter systems are tested to confirmthat user and functional requirementshave been met, and in order to assuredata integrity. These, in turn, are drivenby a regulatory need to assure patientsafety and health.

Validation of systems to guaranteeaccuracy, reliability, consistent in-tended performance, and the ability todiscern invalid or altered records shouldbe considered as part of the completelife cycle of a computer system. Thiscycle includes the stages:

• Planning • Specification• Programming • Testing• Commissioning • Documentation• Operation • Monitoring• Modifying • Decommission-

ing

Figure 2. GAMP® ‘V’ model.

Testing GxP Systems


Before a system using a computer isbrought into use, it should be thor-oughly tested and confirmed as beingcapable of achieving the desired re-sults.

Any modification should undergorisk assessment in order to determinethe extent of the required validationand regression testing. Alterations to asystem or to a computer program shouldbe made only in accordance with adefined procedure, which should in-clude provision for validating, check-ing, approving, and implementing thechange. Change control proceduresshould maintain an audit trail thatdocuments time-sequenced develop-ment and modification of any systemsdocumentation.

There are other regulations whichmay impact the testing of GxP sys-tems, e.g., health and safety legisla-tion, environmental control.

Key PrinciplesThis section outlines the key principlesused in the planning and execution ofGxP systems testing and builds uponthe philosophy of risk-based validationestablished in GAMP® 4.

Use of Risk AssessmentGeneral Principle: the scope of test-ing should be determined by a justifiedand documented risk assessment, tak-ing into account both the potential ef-fect on product quality and safety andthe intrinsic risk associated with themethod of implementation.

As defined in GAMP® 4, AppendixM3, risk assessment should be a fun-damental part of the validation pro-cess. This can be used to determine theappropriate nature and scope of test-ing, as described in this article.

The GxP criticality of the require-ments will provide one indication ofthe risk impact. This requires that theuser understand and interpret the ap-plicable GxP regulations. Combinedwith risk likelihood and probability ofdetection, this will provide an indica-tion of risk priority for each require-ment.

Consideration of the supplier auditand the maturity of the supplier’s pro-cesses and product (in terms of history

of compliance with an appropriate qual-ity system, number of existing users,length of time in market, user satisfac-tion, etc.) will indicate an appropriatescope and rigor of testing. For example,in the case of an established supplier,only positive case acceptance testingmay be required. Where the supplierprocesses or product are less mature, itmay be appropriate to conduct addi-tional testing (negative case testing,software module testing, etc.).

The same risk assessment processshould be applied when changes aremade to the system.

Testing and the GAMP® LifeCycleGeneral Principle: testing should becarried out as part of a formal develop-ment life cycle and test cases should bewritten and executed against docu-mented requirements.

It is an assumption within this ar-ticle that the system requirements havebeen adequately defined and docu-mented and that an appropriate devel-opment life cycle is in place. This guid-ance concentrates on the testing re-quired within that life cycle.

GAMP® 4 describes a framework forspecification and qualification from theuser’s perspective.

The terminology used in the GAMP®

‘V’ model is based upon well under-stood industry qualification activitiesand their relationship to hierarchicalspecifications.

From a regulatory perspective, theresponsibility for qualification activi-ties resides with the user. In a practi-cal sense, testing is usually a sharedresponsibility of both the supplier anduser, which should be agreed at thestart of the project.

Figure 3 (based upon GAMP® 4, Fig-ure 8.2) shows the different types oftesting associated with various require-ments and design specifications. Soft-ware module and some levels of inte-gration testing are usually the respon-sibility of the supplier (software ven-dor or system integrator) whereas theuser is usually responsible for the func-tional, acceptance, and performancetesting (qualification).

The use of the terms ‘testing’, ‘veri-fication’, and ‘validation’ has often beena source of confusion and inconsistency.Within this article:

• ‘Testing’ is one form of verificationactivity that usually forms part ofthe validation process.

• ‘Verification’ is used to describe ameans of confirming that one ormore specific requirements havebeen met.

Figure 3. ‘V-model’ framework showing basic specification and test relationships.

Testing GxP Systems


• ‘Validation’ is used to describe anoverall process.

This basic model is further developedfor different software categories in thefollowing sections and these may ex-pand on the basic model shown above.

Testing StrategiesGeneral Principle: testing shouldfollow an agreed test strategy.

There are a number of points thatshould be considered when developinga test strategy. This may be as part ofa separate test strategy document, ormay be included in the Validation Plan.

Test DocumentationThis section discusses the basic testdocumentation requirements, whichare presented diagrammatically in Fig-ure 4. For users, this test documenta-tion, including any separate test sum-mary report, will be summarized in theValidation Report (refer to GAMP® 4,Appendix M7).

Test documentation should be sub-ject to independent review and ap-proval. This is often conducted by arepresentative assigned by the relevantorganization’s quality function in ac-cordance with their procedures.

When determining an appropriatetest strategy, it should be kept in mindthat the content of documents may becombined or included in another docu-ment, e.g., when testing small or simplesystems:

• Although one-to-many relationshipsare shown in the diagram, in a simplesystem, a single document may in-clude the test strategy, test proto-cols/test specifications, and testcases/test scripts, and a single testreport may be all that is required.

• When testing small or simple sys-tems, test inputs, test environmentset-up, and expected results all maybe covered within the test scripts,and a separate definition of testcases may not be required (this usu-ally is the most common approach).

• The test strategy may form part ofanother planning document.

part of the test environment.The test environment should be

made as representative as possible ofthe final system. Differences should bedocumented and assessed for the levelof impact introduced by the differencesto allow additional tests to be plannedfor the final production system or envi-ronment if required.

The test environment should bedocumented and controlled to a level ofdetail that would allow it to be recon-structed if necessary.

Where test hardware/software/data/user accounts are applied to the finalsystem, controls should exist to ensurethat they can either be removed cleanlyor be isolated from use within the finalproduction environment.

Test ExecutionWhen considering test execution, thetest strategy should consider the meth-ods used for manual test execution andthe methods for automated test execu-tion.

When automatic test tools are used,special care should be taken to assurethey are fit for their intended purpose.

Test Results Recording andReviewingThe test strategy also should considerthe recording and review of test re-sults, including the method for record-ing and filing passed and failed tests.The method for documenting, process-ing, and closing down test incidentsand the requirements for test witnessesalso should be addressed in the teststrategy as well as the review of testresults and associated documentation.

The witnessing and reviewing re-quirements should reflect the relativerisk associated with the system ele-ment under test. For example, for test-ing of simple, low risk elements or fortests where actions and results areautomatically captured, a single suit-ably qualified tester signing off a testmay be appropriate provided that thefinal results are independently re-viewed. For complex or critical func-tions, multiple testers from variousbackgrounds (e.g., supplier, engineer,and user production staff) may be ap-propriate.

For larger or more complex systems thefollowing also should be considered:

• Multiple test protocols or specifica-tions may often be required. Eachtest protocol or specification shouldhave an associated test report, andthe test reports may be summarizedinto a test summary report (which isassociated with a test strategy). Italso may be useful to prepare testplans which will define:- location and timing of test phases- resources required for each phase- responsibilities for each phase- format of test references and in-

cident references- planned coverage for each test

phase (against established re-quirements)

• Test cases and scripts may begrouped into test groups or test sets(not shown) for ease of test plan-ning, monitoring, and execution.

• Separate test cases may be preparedfor some tests, which may describe:- complex test data sets- test methods- test input data- test environment set-up- expected results

In this case, test scripts may be usedsolely to document the sequence of ac-tions (test steps) required to conduct aspecific test. One test script may beused as the basis for conducting mul-tiple similar test cases.

For all systems the test strategyshould describe the use of appropriatetest cases and/or test scripts, includingtest objectives(s), necessary pre-requi-sites, steps involved in performing eachtest, data to be recorded, evidence to becollected, and acceptance criteria.

Test EnvironmentThere will typically be a test environ-ment which is separate from the pro-duction environment and these envi-ronments may be separated logically,physically, or chronologically. The teststrategy should consider the hardware,software, test data sets, user accounts,and reference documents that will be

Testing GxP Systems


Test Reporting and SystemHandoverThe test strategy also should definekey considerations when producing testreports and planning to handover thesystem from one test phase to another,including the format of and responsi-bility for final test reports, the methodof and authority for system handover,and any contractual implications.

The method for system handovermay need to consider circumstances inwhich a conditional handover can bemade – for example with workaroundsin place, test incidents still open ortests still to be completed because theyare not possible outside of the finalenvironment.

The method also should be definedfor ensuring that the baseline recordedat the end of one test phase matchesthe baseline recorded at the start of thenext phase (e.g., to ensure that thesoftware at the start of site acceptancetesting is the same as that released atthe end of factory acceptance testing).

Testing in the OperationalPhaseOnce a system has been implemented,there may be a need for future change.Depending on the scope of the change,there may be a requirement for devel-

oping a test strategy to define thescope and rigor of testing.

Test MetricsTest metrics provide various mea-sures of testing. These allow the testprocess to be assessed, and there-fore, they may be adjusted in a man-aged manner. Testing often takes alarge part of the software develop-ment life cycle, and consequently,making it as efficient as possible isimportant for good management. Inthe quality management cycle of:

• Plan• Do• Check• Act

Test metrics provide a check on theeffectiveness of testing.

Test metrics provide informationthat can justify, refine, and/or im-prove the amount and type of testingused or in certain circumstances todefine that the system is of an ac-ceptable quality for release. Themetrics provide information that isfed into risk assessments and allowmanagers to manage the systemqualification process.

Managers should decide what as-pects of the life cycle and testing they

Figure 4. Basic test documentation model.

would like to control and then look forsuitable metrics for those features.

Testing and Hardware/Software CategoriesGeneral principle: as with other vali-dation efforts, the testing strategyshould reflect risk to product quality,patient safety, and data integrity. Thenature of the software and hardware isone factor affecting this risk.

GAMP® Hardware andSoftware CategoriesSince the risk of system failure in-creases with the progression from stan-dard software and hardware to custom(bespoke) software and hardware, aclassification of system elements intocategories can help support a risk-basedtest strategy. Categorization of hard-ware and software, as described inGAMP® 4, Appendix M4, is assumedthroughout the remainder of this ar-ticle.

Hardware and SoftwareMaturityIn deciding the test effort required fora system element, the maturity of thehardware or software also may need tobe taken into account with additionaleffort devoted to test elements that arenot considered ‘industry proven.’ This

Testing GxP Systems


can apply both to the maturity of thestandard elements within a supplier’sproduct and to the maturity of customelements re-used from one applicationto the next.

For example, where a user is unableto accept the standard functionalityoffered by a supplier and requests amodified module, additional testing ofthe differences between the standardoffering and the modified module islikely to be appropriate. Conversely,where a custom module is re-used infurther applications, a reduced test ef-fort is likely to be appropriate.

Testing Responsibilities -Supplier and UserGeneral principle: where possible,users should seek to benefit from sup-plier quality assurance processes andassociated testing.

Where a supplier has been auditedand their quality management systemfound to be acceptable, the user maybenefit from the testing already car-ried out as part of the product develop-ment life cycle. This may reduce theneed for additional testing carried outby the user.

Regardless of the categorization ofthe software acquired by the user, allsoftware would at some stage havebeen written for the first time by thesupplier and could be considered ascustomized code during developmentby the supplier (synonymous withGAMP® software Category 5). There-fore, it is appropriate to consider thesupplier’s development life cycle (andany integral testing) when considering

the scope and nature of testing to beconducted by the User.

Therefore, the testing of the soft-ware (or system) is a combination of:

• testing conducted by the supplierduring basic development of thestandard product

• testing conducted by the supplier(or integrator) during applicationspecific development

• testing conducted by the user

Where there is auditable evidence thatsupplier testing is appropriate to therisk associated with the software orsystem, the user need not repeat suchtesting, as long as an appropriate Sup-plier Audit has been conducted. Thisshould include a review of general sup-plier test activities and a review ofsystem, software, or release specifictesting. User testing should then focus

on customized, configured, and criticalfunctions.

The examples that follow show sup-plier and user development and testactivities and are based on the basictesting ‘V-model’ life cycle shown ear-lier in this article. Some suppliers mayuse alternative development life cyclesother than those based upon the ‘V-model.’

Alternative development life cyclesmay be perfectly acceptable - the im-portant issue is to focus on the purpose,nature, and scope of the suppliers docu-mented test activities and the veracityof the test results. Where these areappropriate to the risk associated withthe users application of the software,these activities need not be repeatedby the user.

The sections below describe the ap-proach to testing various software cat-egories. Note that most systems con-tain multiple categories of software,and the system specific approach totesting may be a combination of thesemodels.

Operating Systems(GAMP® Software Category 1)It is not necessary to specifically testoperating systems, as these are quali-fied as part of the infrastructure andchallenged indirectly by the functionaltesting of associated applications.

Firmware(GAMP® Software Category 2)The creation of firmware typically in-volves code written by the supplier.Therefore, the supplier should gener-

Figure 6. Test framework for GAMP® software category 3.


Testing GxP Systems


ally follow a full product developmentlife cycle (either to the life cycle recom-mended for GAMP® software Category5 or to equivalent standards).

On purchasing an item with embed-ded firmware, the user does not need torepeat testing already carried out bythe supplier, assuming that the sup-plier has a suitable quality manage-ment system in place and that thefirmware is ‘standard’ (rather thandeveloped or modified specifically forthe application).

The application life cycle verifica-tion activities can be limited to verify-ing the firmware version and that theparameters entered give correct opera-tion as defined in the user require-ments.

Standard Software Packages(GAMP® Software Category 3)The creation of standard software pack-ages typically involves code written bythe supplier. Therefore, the suppliershould generally follow a full productdevelopment life cycle (either to thelife cycle recommended for GAMP® soft-ware Category 5 or to equivalent stan-dards).

On purchasing a standard package,the user does not need to repeat testingalready carried out by the supplier,assuming that the supplier has a suit-able quality management system inplace and that the package is ‘stan-dard’ (rather than developed or modi-fied specifically for the application).

The application life cycle test activi-ties can be limited to verifying theinstalled software package version andthat the parameters entered give cor-rect operation as defined in the userrequirements.

Configurable SoftwarePackages(GAMP® Software Category 4)The creation of configurable softwarepackages typically involves code writ-ten by the supplier. Therefore, the sup-plier should generally follow a full prod-uct development life cycle (either to thelife cycle recommended for GAMP®

Category 5 or to equivalent standards).In order to meet the requirements of

the specific user, the software is then

configured by the supplier, a systemsintegrator, or the user.

On purchasing a configurable pack-age, the user does not need to repeattesting already carried out by the sup-plier, assuming that the supplier has asuitable quality management systemin place and that the package is ‘stan-dard’ (rather than developed or modi-fied specifically for the application).

The application life cycle test activi-ties can be limited to those which verifythat the configuration has been cor-rectly implemented such that the over-all system performs as defined in theuser requirements.

It is not usually necessary for theusers to test functions within the sys-tem that the user does not intend toutilize and where:

• Supplier testing adequately dem-onstrates that unused functions donot interact with functions config-ured and utilized by the user.

• Supplier release notes adequatelydescribe the extent of any interac-tions between functions that areand are not utilized by the user.

Custom (Bespoke) Software(GAMP® Software Category 5)Where users (possibly working withtheir suppliers) develop a system thatsolely contains custom software, Fig-ure 8 shows the users life cycle thatwill be followed. Note that because thesystem is not based upon a standardsupplier product, there is no supplier’slife cycle to be considered.

In the more likely case where cus-tom development is required to modifyor customize a supplier’s standard prod-uct for use in a specific application, thefull development and testing life cycleneeds to be followed for custom modifi-cations. In addition, the suppliers stan-dard configurable software also willneed to be tested as for GAMP® soft-ware Category 4.

Figure 9 looks at the development ofcustom modules to be added onto astandard product (for example to pro-vide an interface to a separate thirdparty software product). The life cyclesgiven above can be followed for testing

the standard package, but the custommodules require a full developmentand test life cycle of their own.

User ConsiderationsUltimate responsibility for testing andconfirming that the system meets itsrequirements lies with the user. It islikely that the user will be acquiringtheir system from one of two sources:

• Supplier: who is developing a prod-uct or a custom system

• Integrator: who is configuring a sys-tem specifically for the user

An integrator is simply a particulartype of supplier. Their supply is likelyto be based on commercial off-the-shelfproducts which they have configuredfor the particular user application.

Although the role of the integratoris highlighted in certain sections ofthis article, the general use of the termSupplier also includes integrators.

The user should be aware of howtheir application has been developedand of the methodologies adopted bythe supplier and associated testing.

Depending on whether a supplier isdeveloping a custom system or an inte-grator is developing a system based onconfigurable software, the user shouldconfirm the approach being adopted.Custom development should follow adevelopment life cycle appropriate forGAMP® software Category 5. Specifi-cation and configuration of commer-cial-off-the-shelf software should fol-low a development life cycle appropri-ate for GAMP® software Category 4.

General ConsiderationsThe user can minimize the level oftesting required by:

• avoiding unnecessary customiza-tion, e.g., by modifying the businessprocess, if this is practical, to matchan off-the-shelf application

• seeking to leverage the testing al-ready executed by the supplier, orpossibly by the user, on identicalsystems or pieces of equipment

Testing GxP Systems


In an ideal situation, user testing maybe reduced to a level which confirmsthat the system meets the user re-quirements and has been tested previ-ously with existing documentary evi-dence for verification.

The user should confirm that thesupplier can meet the requirementsdescribed in this article. This may bedone by an audit process (see GAMP®

4, Appendix M2), but whatever processis used, it should be documented, and

where shortfalls are identified, the userneeds to assess and mitigate the riskson a case-by-case basis.

In cases where suppliers do not havea defined methodology for the testingof their systems, users should consideradditional testing. Where additionaltesting does not appropriately miti-gate the users risks, it may be appro-priate that alternative products and/orsupplier be sought.

Users should encourage suppliers

to address any shortcomings in theirtesting processes and documentationin a systematic manner, possibly aspart of a program of continuous im-provement under a registered qualitysystem such as ISO 9001:2000, TickIT,Software CMM (or CMMI), or an ap-propriate testing maturity model.

The discussion of supplier processmaturity below, refers to a good trackrecord within the healthcare industryand a history of compliance with anappropriate quality system. Productmaturity refers to a history of goodproduct quality with a high level ofcustomer satisfaction in the healthcareindustry.

Alternatively, any deficiencies maybe addressed in a one-off product or ona project basis as part of a plan ofcorrective actions agreed between theuser and supplier, and this may in-clude additional user testing.

Users may consider that, where thesystems are of a highly critical nature,it may be appropriate that correctiveactions to the supplier’s quality systemhave a suitable contractual basis.Should suppliers then fail to conduct ordocument appropriate testing agreedunder such a contract, users may beable to reclaim the cost of additionaluser testing.

Products that are widely used in thehealthcare industry are, generally, con-sidered to be lower risk likelihood thannew products or those developed forgeneral markets and used infrequentlyin the healthcare industry.

Suppliers who are experienced inthe industry pose a lower risk likeli-hood due to their greater understand-ing of regulatory requirements and therisk associated with certain user prod-uct profiles.

Users should seek to purchase lowrisk products and to do business withlow risk suppliers. Market review andan initial Supplier Audit should estab-lish the relative maturity of a productand/or supplier processes and thisshould include a consideration of sup-plier testing.

Regardless of the maturity of theproduct or the supplier processes, thesame level of testing should be con-ducted by all suppliers and this shouldFigure 8. Test framework for GAMP® software category 5.


Testing GxP Systems


be appropriate to the user’s determina-tion of risk. This level of testing shouldalready be in place with an establishedsupplier and this can be assured byinitial Supplier Audit and routine sur-veillance audits.

Users may choose to select productswhich fall into the high/medium riskcategory, but this will usually result inan increase in audit or user testing toensure that the increased risk poten-tial is appropriately addressed (seeGAMP® 4, Appendix M2 for furtherdetails).

Agreeing Roles andResponsibilitiesWhere the Supplier Audit identifiesshortcomings, the user can define ap-propriate actions to mitigate the risks.These would often be detailed withinthe contract or project plan as agreedby all parties. The level of testing andresponsibilities required from both thesupplier and user depend on the cat-egory of system being supplied.

In cases where supplier and productmaturity indicate a low level of risk,the user may determine that they onlyneed to be involved in the execution of

User Acceptance testing with all othertesting conducted by the supplier. Incases where supplier and product ma-turity indicate a high level of risk, theuser may determine that they may alsoneed to witness supplier testing, con-duct additional negative case testing,or repeat poorly documented suppliertesting.

General guidance on appropriateuser/supplier responsibilities for test-ing different software categories is de-scribed under Key Principles in thisarticle.

Determining Appropriate TestEvidenceUsers should define what level of testevidence should be in place before thesystem can be considered to be vali-dated, and how long that evidenceshould be available. The value of suchtest evidence changes over time andthe retention period and the appropri-ate level of test evidence to be retainedcan be determined by risk assessment.

The user should build on evidence oftesting provided by the supplier andaim not to duplicate test evidence.Where supplier test evidence is relied

upon to support the validation of thesystem, users should either requestcopies of the supplier test evidence, orshould assure themselves that suppli-ers test evidence will be retained andavailable for the necessary period. Thismay be assured as part of the SupplierAudit, or specific contractual terms.

Supplier (Integrator)Considerations

The nature of the healthcare industryrequires that systems are developed,documented, and tested following goodengineering practices. Suppliers shouldseek to develop and supply systems inaccordance with a defined methodol-ogy such as that described in GAMP® 4.Good quality testing conducted by thesupplier is likely to allow reduced test-ing by the User.

Supplier AuditFor systems containing software Cat-egory 4 or 5 software (or highly criticalsoftware of Category 2 or 3), it is usualfor a user to carry out an audit of thesupplier’s quality system. The suppliershould make themselves aware of theareas likely to be covered by that audit(see GAMP® 4, Appendix M2, for ex-ample). Being aware of the require-ments and preparing for the audit willassist both parties in determining anyshortfalls and where specific remedialactions or testing may be required. Theaudit may be an important step indeveloping a long term relationshipbetween the supplier and the user.

Use of Third Party ProductsWhere the supplier makes use of thirdparty products at any stage of theirproduct development they should con-sider the quality of their own suppliersand their suppliers’ products whendetermining an appropriate level oftesting. The GAMP® GPG: Testing ofGxP Systems provides assistance tousers in the healthcare industry as tohow they should approach the testingof supplied systems. The same ap-proaches need to be adopted by suppli-ers when they make use of third partyproducts.

Suppliers should be in a position toverify that products they use have beenFigure 9. Test framework for GAMP® software category 4 and 5 (combined).

Testing GxP Systems


developed using good engineering prac-tices and that they have taken all pos-sible measures to ensure this. Thismay involve, but not be limited to:

• Audits of developers of the thirdparty products, this may be re-stricted to a postal audit, but con-sideration should be given to carry-ing out a full audit and if not neededthen the reasons for this docu-mented.

• Specific testing of applications ofthese products, e.g., where specificconfigurations of automated testtools are used these should be testedand produce documentary evidencethat the test tool does what it issupposed to do

• Where third party products are con-sidered to be a “widely used indus-try standard” then suitable evidenceto this should be available.

Similar to users, integrators shouldseek to leverage the testing alreadyexecuted by their supplier(s), or test-ing conducted by themselves on identi-cal systems or pieces of equipment.

Contractual IssuesTesting is often a milestone linked to astage payment. Key points for success

include:

• Agree and document early in theproject the stages and scope of thetesting required – (an assessment ofthe critical functions and risk toend-user can assist with this).

• Ensure the test script links back tothe design specifications (possiblyvia the use of a traceability matrix).

• Involve the supplier and user per-sonnel in the review and approval oftest scripts that are relevant to them.This should ensure that all partiesunderstand the test objectives andshould limit the effects of subse-quent changes.

• Ensure all equipment, includingspare parts, are available prior tothe commencement of testing.

• Ensure that time planned for docu-ment reviews factors in the expectedsize and complexity of the item to bereviewed.

• Ensure all personnel are availablewhen required, including systemdevelopers, in case of a deviationoccurring which requires a changeto the system.

• Prepare contingency and recoveryplans.

On completion of the testing, agree-ment on any outstanding actions ordeviations should be reached betweenthe supplier and user in order for theproject to progress to the next stage.

Content of the SIGGuidelines

The GAMP® GPG: Testing of GxP Sys-tems begins with material included inthis article. Key principles are furtherconsidered and additional practicaladvice and guidance in the planningand execution of such testing are pro-vided. This includes consideration of:

• Testing Policies

• Test Planning and Test Manage-ment

• Test Protocols, Cases, and Scripts

• Test Environments

• Test Execution

• Test Results – Recording and Re-viewing

• Test Reporting and SystemHandover

• Testing in the Operational Phase

Case studies are provided, which applythe key principles to:

• Process Automation Systems

• Configurable IT Systems

• Analytical Instruments

• Desktop Applications

• Infrastructure and Interfaces

Templates and examples are includedthat should allow the less experiencedreader to quickly develop a series ofappropriate documents, suitable forplanning, executing, and reporting onthe testing of a GxP system.

Figure 10. Supplier and product maturity model.

Testing GxP Systems


GlossaryThe following provides a definition ofthe specific terms used in this article.The source is listed at the end of eachdefinition.

Acceptance Criteria - The criteriathat a system or component must sat-isfy in order to be accepted by a user,customer, or other authorized entity.(GAMP® 4, IEEE)

Acceptance Test - Formal testingconducted to determine whether or nota system satisfies its acceptance crite-ria and to enable the customer to deter-mine whether or not to accept the sys-tem. See also Factory Acceptance Test(FAT), Site Acceptance Test (SAT).(GAMP® 4, IEEE)

Firmware - The combination of hard-ware device and computer instructionsand data that reside as read-only soft-ware on that device. (IEEE)

Functional Testing - Testing thatignores the internal mechanism of asystem or component and focuses solelyon the outputs generated in responseto selected input and execution condi-tions. Also known as black box testing.(GAMP® 4, IEEE)

Hardware - (1) Physical equipmentused to process, store, or transmit com-puter programs or data.(2) Physicalequipment used in data processing, asopposed to programs, procedures, rules,and associated documentation. (IEEE)

Hardware Testing - Testing carriedout to verify correct operation of sys-tem hardware independent of any cus-tom application software.

Installation Qualification (IQ) -Documented verification that a systemis installed according to written andpre-approved specifications. (GAMP®

4, PDA)

Integration - The process of combin-ing software components, hardwarecomponents, or both into an overallsystem.Sometimes described as soft-ware integration and system integra-tion respectively. (IEEE)

Integration Testing - (1) Testing inwhich software components, hardwarecomponents, or both are combined andtested to evaluate the interaction be-tween them.(2) An orderly progressionof testing of incremental pieces of thesoftware program in which softwareelements, hardware elements, or bothare combined and tested until the en-tire system has been integrated to showcompliance with the program designed,and capabilities and requirements ofthe system. (IEEE)

Module Testing - Testing of an indi-vidual hardware or software compo-nents or groups of related components.(IEEE)

Negative Testing - Testing aimed atshowing that software does not work.(BCS)

Operational Qualification (OQ) -Documented verification that a systemoperates according to written and pre-approved specifications throughout allspecified operating ranges. (GAMP® 4,PDA)

Performance Qualification (PQ) -Documented verification that a systemis capable of performing or controllingthe activities of the processes it is re-quired to perform or control, accordingto written and pre-approved specifica-tions, while operating in its specifiedoperating environment. (GAMP® 4,PDA)

Positive Testing - Testing aimed atshowing that software does meet thedefined requirements.

Qualification - The process to demon-strate the ability to fulfill specifiedrequirements. (GAMP® 4, ISO)

Software - Computer programs, pro-cedures, and associated documentationand data pertaining to the operation ofa computer system. (IEEE)

System Testing - Testing conductedon a complete, integrated system toevaluate the systems compliance withits specified requirements. (IEEE)

Test - (1) An activity in which a systemor component is executed under spe-cific conditions, the results are observedor recorded, and an evaluation is madeof some aspect of the system or compo-nent. (2) Determination of one or morecharacteristics according to a proce-dure. (GAMP® 4, IEEE), (GAMP® 4,ISO)

Test Case - A set of test inputs, execu-tion conditions, and expected resultsdeveloped for a particular objective,such as to exercise a particular pro-gram path or to verify compliance witha specific requirement. (GAMP® 4,IEEE)

Test Plan - A document describing thescope, approach, resources, and sched-ule of intended test activities. It iden-tifies test items, the features to betested, the testing tasks, who will doeach task, and any risks requiring con-tingency planning. (GAMP® 4, IEEE)

Test Procedure - Detailed instruc-tions for the set-up, execution, andevaluation of results for a given testcase. (GAMP® 4, IEEE)

Test Script - Documentation thatspecifies a sequence of actions for theexecution of a test. (IEEE)

Unit Testing - Testing of individualhardware or software units or groupsof related units. (IEEE)

Validation - Establishing documentedevidence which provides a high degreeof assurance that a specific process willconsistently produce a product meet-ing its pre-determined specificationsand quality attributes. (GAMP® 4, FDA)

Verification - Confirmation, throughthe provision of objective evidence thatspecified requirements have been ful-filled. (GAMP® 4, ISO)